Files
deer-flow/backend/tests/test_sandbox_tools_security.py

426 lines
18 KiB
Python
Raw Permalink Normal View History

feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
from pathlib import Path
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
from unittest.mock import patch
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
import pytest
refactor: split backend into harness (deerflow.*) and app (app.*) (#1131) * refactor: extract shared utils to break harness→app cross-layer imports Move _validate_skill_frontmatter to src/skills/validation.py and CONVERTIBLE_EXTENSIONS + convert_file_to_markdown to src/utils/file_conversion.py. This eliminates the two reverse dependencies from client.py (harness layer) into gateway/routers/ (app layer), preparing for the harness/app package split. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: split backend/src into harness (deerflow.*) and app (app.*) Physically split the monolithic backend/src/ package into two layers: - **Harness** (`packages/harness/deerflow/`): publishable agent framework package with import prefix `deerflow.*`. Contains agents, sandbox, tools, models, MCP, skills, config, and all core infrastructure. - **App** (`app/`): unpublished application code with import prefix `app.*`. Contains gateway (FastAPI REST API) and channels (IM integrations). Key changes: - Move 13 harness modules to packages/harness/deerflow/ via git mv - Move gateway + channels to app/ via git mv - Rename all imports: src.* → deerflow.* (harness) / app.* (app layer) - Set up uv workspace with deerflow-harness as workspace member - Update langgraph.json, config.example.yaml, all scripts, Docker files - Add build-system (hatchling) to harness pyproject.toml - Add PYTHONPATH=. to gateway startup commands for app.* resolution - Update ruff.toml with known-first-party for import sorting - Update all documentation to reflect new directory structure Boundary rule enforced: harness code never imports from app. All 429 tests pass. Lint clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: add harness→app boundary check test and update docs Add test_harness_boundary.py that scans all Python files in packages/harness/deerflow/ and fails if any `from app.*` or `import app.*` statement is found. This enforces the architectural rule that the harness layer never depends on the app layer. Update CLAUDE.md to document the harness/app split architecture, import conventions, and the boundary enforcement test. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: add config versioning with auto-upgrade on startup When config.example.yaml schema changes, developers' local config.yaml files can silently become outdated. This adds a config_version field and auto-upgrade mechanism so breaking changes (like src.* → deerflow.* renames) are applied automatically before services start. - Add config_version: 1 to config.example.yaml - Add startup version check warning in AppConfig.from_file() - Add scripts/config-upgrade.sh with migration registry for value replacements - Add `make config-upgrade` target - Auto-run config-upgrade in serve.sh and start-daemon.sh before starting services - Add config error hints in service failure messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix comments * fix: update src.* import in test_sandbox_tools_security to deerflow.* Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle empty config and search parent dirs for config.example.yaml Address Copilot review comments on PR #1131: - Guard against yaml.safe_load() returning None for empty config files - Search parent directories for config.example.yaml instead of only looking next to config.yaml, fixing detection in common setups Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct skills root path depth and config_version type coercion - loader.py: fix get_skills_root_path() to use 5 parent levels (was 3) after harness split, file lives at packages/harness/deerflow/skills/ so parent×3 resolved to backend/packages/harness/ instead of backend/ - app_config.py: coerce config_version to int() before comparison in _check_config_version() to prevent TypeError when YAML stores value as string (e.g. config_version: "1") - tests: add regression tests for both fixes Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: update test imports from src.* to deerflow.*/app.* after harness refactor Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-14 22:55:52 +08:00
from deerflow.sandbox.tools import (
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
VIRTUAL_PATH_PREFIX,
feat(harness): integration ACP agent tool (#1344) * refactor: extract shared utils to break harness→app cross-layer imports Move _validate_skill_frontmatter to src/skills/validation.py and CONVERTIBLE_EXTENSIONS + convert_file_to_markdown to src/utils/file_conversion.py. This eliminates the two reverse dependencies from client.py (harness layer) into gateway/routers/ (app layer), preparing for the harness/app package split. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: split backend/src into harness (deerflow.*) and app (app.*) Physically split the monolithic backend/src/ package into two layers: - **Harness** (`packages/harness/deerflow/`): publishable agent framework package with import prefix `deerflow.*`. Contains agents, sandbox, tools, models, MCP, skills, config, and all core infrastructure. - **App** (`app/`): unpublished application code with import prefix `app.*`. Contains gateway (FastAPI REST API) and channels (IM integrations). Key changes: - Move 13 harness modules to packages/harness/deerflow/ via git mv - Move gateway + channels to app/ via git mv - Rename all imports: src.* → deerflow.* (harness) / app.* (app layer) - Set up uv workspace with deerflow-harness as workspace member - Update langgraph.json, config.example.yaml, all scripts, Docker files - Add build-system (hatchling) to harness pyproject.toml - Add PYTHONPATH=. to gateway startup commands for app.* resolution - Update ruff.toml with known-first-party for import sorting - Update all documentation to reflect new directory structure Boundary rule enforced: harness code never imports from app. All 429 tests pass. Lint clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: add harness→app boundary check test and update docs Add test_harness_boundary.py that scans all Python files in packages/harness/deerflow/ and fails if any `from app.*` or `import app.*` statement is found. This enforces the architectural rule that the harness layer never depends on the app layer. Update CLAUDE.md to document the harness/app split architecture, import conventions, and the boundary enforcement test. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: add config versioning with auto-upgrade on startup When config.example.yaml schema changes, developers' local config.yaml files can silently become outdated. This adds a config_version field and auto-upgrade mechanism so breaking changes (like src.* → deerflow.* renames) are applied automatically before services start. - Add config_version: 1 to config.example.yaml - Add startup version check warning in AppConfig.from_file() - Add scripts/config-upgrade.sh with migration registry for value replacements - Add `make config-upgrade` target - Auto-run config-upgrade in serve.sh and start-daemon.sh before starting services - Add config error hints in service failure messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix comments * fix: update src.* import in test_sandbox_tools_security to deerflow.* Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle empty config and search parent dirs for config.example.yaml Address Copilot review comments on PR #1131: - Guard against yaml.safe_load() returning None for empty config files - Search parent directories for config.example.yaml instead of only looking next to config.yaml, fixing detection in common setups Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct skills root path depth and config_version type coercion - loader.py: fix get_skills_root_path() to use 5 parent levels (was 3) after harness split, file lives at packages/harness/deerflow/skills/ so parent×3 resolved to backend/packages/harness/ instead of backend/ - app_config.py: coerce config_version to int() before comparison in _check_config_version() to prevent TypeError when YAML stores value as string (e.g. config_version: "1") - tests: add regression tests for both fixes Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: update test imports from src.* to deerflow.*/app.* after harness refactor Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(harness): add tool-first ACP agent invocation (#37) * feat(harness): add tool-first ACP agent invocation * build(harness): make ACP dependency required * fix(harness): address ACP review feedback * feat(harness): decouple ACP agent workspace from thread data ACP agents (codex, claude-code) previously used per-thread workspace directories, causing path resolution complexity and coupling task execution to DeerFlow's internal thread data layout. This change: - Replace _resolve_cwd() with a fixed _get_work_dir() that always uses {base_dir}/acp-workspace/, eliminating virtual path translation and thread_id lookups - Introduce /mnt/acp-workspace virtual path for lead agent read-only access to ACP agent output files (same pattern as /mnt/skills) - Add security guards: read-only validation, path traversal prevention, command path allowlisting, and output masking for acp-workspace - Update system prompt and tool description to guide LLM: send self-contained tasks to ACP agents, copy results via /mnt/acp-workspace - Add 11 new security tests for ACP workspace path handling Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(prompt): inject ACP section only when ACP agents are configured The ACP agent guidance in the system prompt is now conditionally built by _build_acp_section(), which checks get_acp_agents() and returns an empty string when no ACP agents are configured. This avoids polluting the prompt with irrelevant instructions for users who don't use ACP. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix lint * fix(harness): address Copilot review comments on sandbox path handling and ACP tool - local_sandbox: fix path-segment boundary bug in _resolve_path (== or startswith +"/") and add lookahead in _resolve_paths_in_command regex to prevent /mnt/skills matching inside /mnt/skills-extra - local_sandbox_provider: replace print() with logger.warning(..., exc_info=True) - invoke_acp_agent_tool: guard getattr(option, "optionId") with None default + continue; move full prompt from INFO to DEBUG level (truncated to 200 chars) - sandbox/tools: fix _get_acp_workspace_host_path docstring to match implementation; remove misleading "read-only" language from validate_local_bash_command_paths Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(acp): thread-isolated workspaces, permission guardrail, and ContextVar registry P1.1 – ACP workspace thread isolation - Add `Paths.acp_workspace_dir(thread_id)` for per-thread paths - `_get_work_dir(thread_id)` in invoke_acp_agent_tool now uses `{base_dir}/threads/{thread_id}/acp-workspace/`; falls back to global workspace when thread_id is absent or invalid - `_invoke` extracts thread_id from `RunnableConfig` via `Annotated[RunnableConfig, InjectedToolArg]` - `sandbox/tools.py`: `_get_acp_workspace_host_path(thread_id)`, `_resolve_acp_workspace_path(path, thread_id)`, and all callers (`replace_virtual_paths_in_command`, `mask_local_paths_in_output`, `ls_tool`, `read_file_tool`) now resolve ACP paths per-thread P1.2 – ACP permission guardrail - New `auto_approve_permissions: bool = False` field in `ACPAgentConfig` - `_build_permission_response(options, *, auto_approve: bool)` now defaults to deny; only approves when `auto_approve=True` - Document field in `config.example.yaml` P2 – Deferred tool registry race condition - Replace module-level `_registry` global with `contextvars.ContextVar` - Each asyncio request context gets its own registry; worker threads inherit the context automatically via `loop.run_in_executor` - Expose `get_deferred_registry` / `set_deferred_registry` / `reset_deferred_registry` helpers Tests: 831 pass (57 for affected modules, 3 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(sandbox): mount /mnt/acp-workspace in docker sandbox container The AioSandboxProvider was not mounting the ACP workspace into the sandbox container, so /mnt/acp-workspace was inaccessible when the lead agent tried to read ACP results in docker mode. Changes: - `ensure_thread_dirs`: also create `acp-workspace/` (chmod 0o777) so the directory exists before the sandbox container starts — required for Docker volume mounts - `_get_thread_mounts`: add read-only `/mnt/acp-workspace` mount using the per-thread host path (`host_paths.acp_workspace_dir(thread_id)`) - Update stale CLAUDE.md description (was "fixed global workspace") Tests: `test_aio_sandbox_provider.py` (4 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(lint): remove unused imports in test_aio_sandbox_provider Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix config --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-26 14:20:18 +08:00
_is_acp_workspace_path,
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
_is_skills_path,
_reject_path_traversal,
feat(harness): integration ACP agent tool (#1344) * refactor: extract shared utils to break harness→app cross-layer imports Move _validate_skill_frontmatter to src/skills/validation.py and CONVERTIBLE_EXTENSIONS + convert_file_to_markdown to src/utils/file_conversion.py. This eliminates the two reverse dependencies from client.py (harness layer) into gateway/routers/ (app layer), preparing for the harness/app package split. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: split backend/src into harness (deerflow.*) and app (app.*) Physically split the monolithic backend/src/ package into two layers: - **Harness** (`packages/harness/deerflow/`): publishable agent framework package with import prefix `deerflow.*`. Contains agents, sandbox, tools, models, MCP, skills, config, and all core infrastructure. - **App** (`app/`): unpublished application code with import prefix `app.*`. Contains gateway (FastAPI REST API) and channels (IM integrations). Key changes: - Move 13 harness modules to packages/harness/deerflow/ via git mv - Move gateway + channels to app/ via git mv - Rename all imports: src.* → deerflow.* (harness) / app.* (app layer) - Set up uv workspace with deerflow-harness as workspace member - Update langgraph.json, config.example.yaml, all scripts, Docker files - Add build-system (hatchling) to harness pyproject.toml - Add PYTHONPATH=. to gateway startup commands for app.* resolution - Update ruff.toml with known-first-party for import sorting - Update all documentation to reflect new directory structure Boundary rule enforced: harness code never imports from app. All 429 tests pass. Lint clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: add harness→app boundary check test and update docs Add test_harness_boundary.py that scans all Python files in packages/harness/deerflow/ and fails if any `from app.*` or `import app.*` statement is found. This enforces the architectural rule that the harness layer never depends on the app layer. Update CLAUDE.md to document the harness/app split architecture, import conventions, and the boundary enforcement test. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: add config versioning with auto-upgrade on startup When config.example.yaml schema changes, developers' local config.yaml files can silently become outdated. This adds a config_version field and auto-upgrade mechanism so breaking changes (like src.* → deerflow.* renames) are applied automatically before services start. - Add config_version: 1 to config.example.yaml - Add startup version check warning in AppConfig.from_file() - Add scripts/config-upgrade.sh with migration registry for value replacements - Add `make config-upgrade` target - Auto-run config-upgrade in serve.sh and start-daemon.sh before starting services - Add config error hints in service failure messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix comments * fix: update src.* import in test_sandbox_tools_security to deerflow.* Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle empty config and search parent dirs for config.example.yaml Address Copilot review comments on PR #1131: - Guard against yaml.safe_load() returning None for empty config files - Search parent directories for config.example.yaml instead of only looking next to config.yaml, fixing detection in common setups Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct skills root path depth and config_version type coercion - loader.py: fix get_skills_root_path() to use 5 parent levels (was 3) after harness split, file lives at packages/harness/deerflow/skills/ so parent×3 resolved to backend/packages/harness/ instead of backend/ - app_config.py: coerce config_version to int() before comparison in _check_config_version() to prevent TypeError when YAML stores value as string (e.g. config_version: "1") - tests: add regression tests for both fixes Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: update test imports from src.* to deerflow.*/app.* after harness refactor Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(harness): add tool-first ACP agent invocation (#37) * feat(harness): add tool-first ACP agent invocation * build(harness): make ACP dependency required * fix(harness): address ACP review feedback * feat(harness): decouple ACP agent workspace from thread data ACP agents (codex, claude-code) previously used per-thread workspace directories, causing path resolution complexity and coupling task execution to DeerFlow's internal thread data layout. This change: - Replace _resolve_cwd() with a fixed _get_work_dir() that always uses {base_dir}/acp-workspace/, eliminating virtual path translation and thread_id lookups - Introduce /mnt/acp-workspace virtual path for lead agent read-only access to ACP agent output files (same pattern as /mnt/skills) - Add security guards: read-only validation, path traversal prevention, command path allowlisting, and output masking for acp-workspace - Update system prompt and tool description to guide LLM: send self-contained tasks to ACP agents, copy results via /mnt/acp-workspace - Add 11 new security tests for ACP workspace path handling Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(prompt): inject ACP section only when ACP agents are configured The ACP agent guidance in the system prompt is now conditionally built by _build_acp_section(), which checks get_acp_agents() and returns an empty string when no ACP agents are configured. This avoids polluting the prompt with irrelevant instructions for users who don't use ACP. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix lint * fix(harness): address Copilot review comments on sandbox path handling and ACP tool - local_sandbox: fix path-segment boundary bug in _resolve_path (== or startswith +"/") and add lookahead in _resolve_paths_in_command regex to prevent /mnt/skills matching inside /mnt/skills-extra - local_sandbox_provider: replace print() with logger.warning(..., exc_info=True) - invoke_acp_agent_tool: guard getattr(option, "optionId") with None default + continue; move full prompt from INFO to DEBUG level (truncated to 200 chars) - sandbox/tools: fix _get_acp_workspace_host_path docstring to match implementation; remove misleading "read-only" language from validate_local_bash_command_paths Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(acp): thread-isolated workspaces, permission guardrail, and ContextVar registry P1.1 – ACP workspace thread isolation - Add `Paths.acp_workspace_dir(thread_id)` for per-thread paths - `_get_work_dir(thread_id)` in invoke_acp_agent_tool now uses `{base_dir}/threads/{thread_id}/acp-workspace/`; falls back to global workspace when thread_id is absent or invalid - `_invoke` extracts thread_id from `RunnableConfig` via `Annotated[RunnableConfig, InjectedToolArg]` - `sandbox/tools.py`: `_get_acp_workspace_host_path(thread_id)`, `_resolve_acp_workspace_path(path, thread_id)`, and all callers (`replace_virtual_paths_in_command`, `mask_local_paths_in_output`, `ls_tool`, `read_file_tool`) now resolve ACP paths per-thread P1.2 – ACP permission guardrail - New `auto_approve_permissions: bool = False` field in `ACPAgentConfig` - `_build_permission_response(options, *, auto_approve: bool)` now defaults to deny; only approves when `auto_approve=True` - Document field in `config.example.yaml` P2 – Deferred tool registry race condition - Replace module-level `_registry` global with `contextvars.ContextVar` - Each asyncio request context gets its own registry; worker threads inherit the context automatically via `loop.run_in_executor` - Expose `get_deferred_registry` / `set_deferred_registry` / `reset_deferred_registry` helpers Tests: 831 pass (57 for affected modules, 3 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(sandbox): mount /mnt/acp-workspace in docker sandbox container The AioSandboxProvider was not mounting the ACP workspace into the sandbox container, so /mnt/acp-workspace was inaccessible when the lead agent tried to read ACP results in docker mode. Changes: - `ensure_thread_dirs`: also create `acp-workspace/` (chmod 0o777) so the directory exists before the sandbox container starts — required for Docker volume mounts - `_get_thread_mounts`: add read-only `/mnt/acp-workspace` mount using the per-thread host path (`host_paths.acp_workspace_dir(thread_id)`) - Update stale CLAUDE.md description (was "fixed global workspace") Tests: `test_aio_sandbox_provider.py` (4 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(lint): remove unused imports in test_aio_sandbox_provider Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix config --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-26 14:20:18 +08:00
_resolve_acp_workspace_path,
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
_resolve_and_validate_user_data_path,
_resolve_skills_path,
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
mask_local_paths_in_output,
replace_virtual_path,
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
replace_virtual_paths_in_command,
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
validate_local_bash_command_paths,
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
validate_local_tool_path,
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
)
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
_THREAD_DATA = {
"workspace_path": "/tmp/deer-flow/threads/t1/user-data/workspace",
"uploads_path": "/tmp/deer-flow/threads/t1/user-data/uploads",
"outputs_path": "/tmp/deer-flow/threads/t1/user-data/outputs",
}
# ---------- replace_virtual_path ----------
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
def test_replace_virtual_path_maps_virtual_root_and_subpaths() -> None:
feat(harness): integration ACP agent tool (#1344) * refactor: extract shared utils to break harness→app cross-layer imports Move _validate_skill_frontmatter to src/skills/validation.py and CONVERTIBLE_EXTENSIONS + convert_file_to_markdown to src/utils/file_conversion.py. This eliminates the two reverse dependencies from client.py (harness layer) into gateway/routers/ (app layer), preparing for the harness/app package split. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: split backend/src into harness (deerflow.*) and app (app.*) Physically split the monolithic backend/src/ package into two layers: - **Harness** (`packages/harness/deerflow/`): publishable agent framework package with import prefix `deerflow.*`. Contains agents, sandbox, tools, models, MCP, skills, config, and all core infrastructure. - **App** (`app/`): unpublished application code with import prefix `app.*`. Contains gateway (FastAPI REST API) and channels (IM integrations). Key changes: - Move 13 harness modules to packages/harness/deerflow/ via git mv - Move gateway + channels to app/ via git mv - Rename all imports: src.* → deerflow.* (harness) / app.* (app layer) - Set up uv workspace with deerflow-harness as workspace member - Update langgraph.json, config.example.yaml, all scripts, Docker files - Add build-system (hatchling) to harness pyproject.toml - Add PYTHONPATH=. to gateway startup commands for app.* resolution - Update ruff.toml with known-first-party for import sorting - Update all documentation to reflect new directory structure Boundary rule enforced: harness code never imports from app. All 429 tests pass. Lint clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: add harness→app boundary check test and update docs Add test_harness_boundary.py that scans all Python files in packages/harness/deerflow/ and fails if any `from app.*` or `import app.*` statement is found. This enforces the architectural rule that the harness layer never depends on the app layer. Update CLAUDE.md to document the harness/app split architecture, import conventions, and the boundary enforcement test. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: add config versioning with auto-upgrade on startup When config.example.yaml schema changes, developers' local config.yaml files can silently become outdated. This adds a config_version field and auto-upgrade mechanism so breaking changes (like src.* → deerflow.* renames) are applied automatically before services start. - Add config_version: 1 to config.example.yaml - Add startup version check warning in AppConfig.from_file() - Add scripts/config-upgrade.sh with migration registry for value replacements - Add `make config-upgrade` target - Auto-run config-upgrade in serve.sh and start-daemon.sh before starting services - Add config error hints in service failure messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix comments * fix: update src.* import in test_sandbox_tools_security to deerflow.* Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle empty config and search parent dirs for config.example.yaml Address Copilot review comments on PR #1131: - Guard against yaml.safe_load() returning None for empty config files - Search parent directories for config.example.yaml instead of only looking next to config.yaml, fixing detection in common setups Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct skills root path depth and config_version type coercion - loader.py: fix get_skills_root_path() to use 5 parent levels (was 3) after harness split, file lives at packages/harness/deerflow/skills/ so parent×3 resolved to backend/packages/harness/ instead of backend/ - app_config.py: coerce config_version to int() before comparison in _check_config_version() to prevent TypeError when YAML stores value as string (e.g. config_version: "1") - tests: add regression tests for both fixes Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: update test imports from src.* to deerflow.*/app.* after harness refactor Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(harness): add tool-first ACP agent invocation (#37) * feat(harness): add tool-first ACP agent invocation * build(harness): make ACP dependency required * fix(harness): address ACP review feedback * feat(harness): decouple ACP agent workspace from thread data ACP agents (codex, claude-code) previously used per-thread workspace directories, causing path resolution complexity and coupling task execution to DeerFlow's internal thread data layout. This change: - Replace _resolve_cwd() with a fixed _get_work_dir() that always uses {base_dir}/acp-workspace/, eliminating virtual path translation and thread_id lookups - Introduce /mnt/acp-workspace virtual path for lead agent read-only access to ACP agent output files (same pattern as /mnt/skills) - Add security guards: read-only validation, path traversal prevention, command path allowlisting, and output masking for acp-workspace - Update system prompt and tool description to guide LLM: send self-contained tasks to ACP agents, copy results via /mnt/acp-workspace - Add 11 new security tests for ACP workspace path handling Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(prompt): inject ACP section only when ACP agents are configured The ACP agent guidance in the system prompt is now conditionally built by _build_acp_section(), which checks get_acp_agents() and returns an empty string when no ACP agents are configured. This avoids polluting the prompt with irrelevant instructions for users who don't use ACP. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix lint * fix(harness): address Copilot review comments on sandbox path handling and ACP tool - local_sandbox: fix path-segment boundary bug in _resolve_path (== or startswith +"/") and add lookahead in _resolve_paths_in_command regex to prevent /mnt/skills matching inside /mnt/skills-extra - local_sandbox_provider: replace print() with logger.warning(..., exc_info=True) - invoke_acp_agent_tool: guard getattr(option, "optionId") with None default + continue; move full prompt from INFO to DEBUG level (truncated to 200 chars) - sandbox/tools: fix _get_acp_workspace_host_path docstring to match implementation; remove misleading "read-only" language from validate_local_bash_command_paths Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(acp): thread-isolated workspaces, permission guardrail, and ContextVar registry P1.1 – ACP workspace thread isolation - Add `Paths.acp_workspace_dir(thread_id)` for per-thread paths - `_get_work_dir(thread_id)` in invoke_acp_agent_tool now uses `{base_dir}/threads/{thread_id}/acp-workspace/`; falls back to global workspace when thread_id is absent or invalid - `_invoke` extracts thread_id from `RunnableConfig` via `Annotated[RunnableConfig, InjectedToolArg]` - `sandbox/tools.py`: `_get_acp_workspace_host_path(thread_id)`, `_resolve_acp_workspace_path(path, thread_id)`, and all callers (`replace_virtual_paths_in_command`, `mask_local_paths_in_output`, `ls_tool`, `read_file_tool`) now resolve ACP paths per-thread P1.2 – ACP permission guardrail - New `auto_approve_permissions: bool = False` field in `ACPAgentConfig` - `_build_permission_response(options, *, auto_approve: bool)` now defaults to deny; only approves when `auto_approve=True` - Document field in `config.example.yaml` P2 – Deferred tool registry race condition - Replace module-level `_registry` global with `contextvars.ContextVar` - Each asyncio request context gets its own registry; worker threads inherit the context automatically via `loop.run_in_executor` - Expose `get_deferred_registry` / `set_deferred_registry` / `reset_deferred_registry` helpers Tests: 831 pass (57 for affected modules, 3 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(sandbox): mount /mnt/acp-workspace in docker sandbox container The AioSandboxProvider was not mounting the ACP workspace into the sandbox container, so /mnt/acp-workspace was inaccessible when the lead agent tried to read ACP results in docker mode. Changes: - `ensure_thread_dirs`: also create `acp-workspace/` (chmod 0o777) so the directory exists before the sandbox container starts — required for Docker volume mounts - `_get_thread_mounts`: add read-only `/mnt/acp-workspace` mount using the per-thread host path (`host_paths.acp_workspace_dir(thread_id)`) - Update stale CLAUDE.md description (was "fixed global workspace") Tests: `test_aio_sandbox_provider.py` (4 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(lint): remove unused imports in test_aio_sandbox_provider Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix config --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-26 14:20:18 +08:00
assert Path(replace_virtual_path("/mnt/user-data/workspace/a.txt", _THREAD_DATA)).as_posix() == "/tmp/deer-flow/threads/t1/user-data/workspace/a.txt"
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
assert Path(replace_virtual_path("/mnt/user-data", _THREAD_DATA)).as_posix() == "/tmp/deer-flow/threads/t1/user-data"
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
# ---------- mask_local_paths_in_output ----------
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_mask_local_paths_in_output_hides_host_paths() -> None:
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
output = "Created: /tmp/deer-flow/threads/t1/user-data/workspace/result.txt"
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
masked = mask_local_paths_in_output(output, _THREAD_DATA)
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
assert "/tmp/deer-flow/threads/t1/user-data" not in masked
assert "/mnt/user-data/workspace/result.txt" in masked
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_mask_local_paths_in_output_hides_skills_host_paths() -> None:
"""Skills host paths in bash output should be masked to virtual paths."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value="/home/user/deer-flow/skills"),
):
output = "Reading: /home/user/deer-flow/skills/public/bootstrap/SKILL.md"
masked = mask_local_paths_in_output(output, _THREAD_DATA)
assert "/home/user/deer-flow/skills" not in masked
assert "/mnt/skills/public/bootstrap/SKILL.md" in masked
# ---------- _reject_path_traversal ----------
def test_reject_path_traversal_blocks_dotdot() -> None:
with pytest.raises(PermissionError, match="path traversal"):
_reject_path_traversal("/mnt/user-data/workspace/../../etc/passwd")
def test_reject_path_traversal_blocks_dotdot_at_start() -> None:
with pytest.raises(PermissionError, match="path traversal"):
_reject_path_traversal("../etc/passwd")
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_reject_path_traversal_blocks_backslash_dotdot() -> None:
with pytest.raises(PermissionError, match="path traversal"):
_reject_path_traversal("/mnt/user-data/workspace\\..\\..\\etc\\passwd")
def test_reject_path_traversal_allows_normal_paths() -> None:
# Should not raise
_reject_path_traversal("/mnt/user-data/workspace/file.txt")
_reject_path_traversal("/mnt/skills/public/bootstrap/SKILL.md")
_reject_path_traversal("/mnt/user-data/workspace/sub/dir/file.py")
# ---------- validate_local_tool_path ----------
def test_validate_local_tool_path_rejects_non_virtual_path() -> None:
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
with pytest.raises(PermissionError, match="Only paths under"):
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
validate_local_tool_path("/Users/someone/config.yaml", _THREAD_DATA)
def test_validate_local_tool_path_rejects_bare_virtual_root() -> None:
"""The bare /mnt/user-data root without trailing slash is not a valid sub-path."""
with pytest.raises(PermissionError, match="Only paths under"):
validate_local_tool_path(VIRTUAL_PATH_PREFIX, _THREAD_DATA)
def test_validate_local_tool_path_allows_user_data_paths() -> None:
# Should not raise — user-data paths are always allowed
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/workspace/file.txt", _THREAD_DATA)
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/uploads/doc.pdf", _THREAD_DATA)
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/outputs/result.csv", _THREAD_DATA)
def test_validate_local_tool_path_allows_user_data_write() -> None:
# read_only=False (default) should still work for user-data paths
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/workspace/file.txt", _THREAD_DATA, read_only=False)
def test_validate_local_tool_path_rejects_traversal_in_user_data() -> None:
"""Path traversal via .. in user-data paths must be rejected."""
with pytest.raises(PermissionError, match="path traversal"):
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/workspace/../../etc/passwd", _THREAD_DATA)
def test_validate_local_tool_path_rejects_traversal_in_skills() -> None:
"""Path traversal via .. in skills paths must be rejected."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
with pytest.raises(PermissionError, match="path traversal"):
validate_local_tool_path("/mnt/skills/../../etc/passwd", _THREAD_DATA, read_only=True)
def test_validate_local_tool_path_rejects_none_thread_data() -> None:
"""Missing thread_data should raise SandboxRuntimeError."""
from deerflow.sandbox.exceptions import SandboxRuntimeError
with pytest.raises(SandboxRuntimeError):
validate_local_tool_path(f"{VIRTUAL_PATH_PREFIX}/workspace/file.txt", None)
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
# ---------- _resolve_skills_path ----------
def test_resolve_skills_path_resolves_correctly() -> None:
"""Skills virtual path should resolve to host path."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value="/home/user/deer-flow/skills"),
):
resolved = _resolve_skills_path("/mnt/skills/public/bootstrap/SKILL.md")
assert resolved == "/home/user/deer-flow/skills/public/bootstrap/SKILL.md"
def test_resolve_skills_path_resolves_root() -> None:
"""Skills container root should resolve to host skills directory."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value="/home/user/deer-flow/skills"),
):
resolved = _resolve_skills_path("/mnt/skills")
assert resolved == "/home/user/deer-flow/skills"
def test_resolve_skills_path_raises_when_not_configured() -> None:
"""Should raise FileNotFoundError when skills directory is not available."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value=None),
):
with pytest.raises(FileNotFoundError, match="Skills directory not available"):
_resolve_skills_path("/mnt/skills/public/bootstrap/SKILL.md")
# ---------- _resolve_and_validate_user_data_path ----------
def test_resolve_and_validate_user_data_path_resolves_correctly(tmp_path: Path) -> None:
"""Resolved path should land inside the correct thread directory."""
workspace = tmp_path / "workspace"
workspace.mkdir()
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
thread_data = {
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
"workspace_path": str(workspace),
"uploads_path": str(tmp_path / "uploads"),
"outputs_path": str(tmp_path / "outputs"),
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
}
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
resolved = _resolve_and_validate_user_data_path("/mnt/user-data/workspace/hello.txt", thread_data)
assert resolved == str(workspace / "hello.txt")
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_resolve_and_validate_user_data_path_blocks_traversal(tmp_path: Path) -> None:
"""Even after resolution, path must stay within allowed roots."""
workspace = tmp_path / "workspace"
workspace.mkdir()
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
thread_data = {
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
"workspace_path": str(workspace),
"uploads_path": str(tmp_path / "uploads"),
"outputs_path": str(tmp_path / "outputs"),
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
}
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
# This path resolves outside the allowed roots
with pytest.raises(PermissionError):
_resolve_and_validate_user_data_path("/mnt/user-data/workspace/../../../etc/passwd", thread_data)
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
# ---------- replace_virtual_paths_in_command ----------
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_replace_virtual_paths_in_command_replaces_skills_paths() -> None:
"""Skills virtual paths in commands should be resolved to host paths."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value="/home/user/deer-flow/skills"),
):
cmd = "cat /mnt/skills/public/bootstrap/SKILL.md"
result = replace_virtual_paths_in_command(cmd, _THREAD_DATA)
assert "/mnt/skills" not in result
assert "/home/user/deer-flow/skills/public/bootstrap/SKILL.md" in result
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_replace_virtual_paths_in_command_replaces_both() -> None:
"""Both user-data and skills paths should be replaced in the same command."""
with (
patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"),
patch("deerflow.sandbox.tools._get_skills_host_path", return_value="/home/user/skills"),
):
cmd = "cat /mnt/skills/public/SKILL.md > /mnt/user-data/workspace/out.txt"
result = replace_virtual_paths_in_command(cmd, _THREAD_DATA)
assert "/mnt/skills" not in result
assert "/mnt/user-data" not in result
assert "/home/user/skills/public/SKILL.md" in result
assert "/tmp/deer-flow/threads/t1/user-data/workspace/out.txt" in result
# ---------- validate_local_bash_command_paths ----------
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_validate_local_bash_command_paths_blocks_host_paths() -> None:
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
with pytest.raises(PermissionError, match="Unsafe absolute paths"):
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
validate_local_bash_command_paths("cat /etc/passwd", _THREAD_DATA)
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
def test_validate_local_bash_command_paths_allows_virtual_and_system_paths() -> None:
validate_local_bash_command_paths(
"/bin/echo ok > /mnt/user-data/workspace/out.txt && cat /dev/null",
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
_THREAD_DATA,
feat(sandbox): harden local file access and mask host paths (#983) * feat(sandbox): harden local file access and mask host paths - enforce local sandbox file tools to only accept /mnt/user-data paths - add path traversal checks against thread workspace/uploads/outputs roots - preserve requested virtual paths in tool error messages (no host path leaks) - mask local absolute paths in bash output back to virtual sandbox paths - update bash tool guidance to prefer thread-local venv + python -m pip - add regression tests for path mapping, masking, and access restrictions Fixes #968 * feat(sandbox): restrict risky absolute paths in local bash commands - validate absolute path usage in local-mode bash commands - allow only /mnt/user-data virtual paths for user data access - keep a small allowlist for system executable/device paths - return clear permission errors for unsafe command paths - add regression tests for bash path validation rules * test(sandbox): add success path test for resolve_local_tool_path (#992) * Initial plan * test(sandbox): add success path test for resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path (#991) * Initial plan * fix(sandbox): reject bare virtual root early with clear error in resolve_local_tool_path Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: WillemJiang <219644+WillemJiang@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> --------- Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2026-03-13 22:38:32 +08:00
)
fix(harness): allow agent read access to /mnt/skills in local sandbox (#1178) * fix(harness): allow agent read access to /mnt/skills in local sandbox Skill files under /mnt/skills/ were blocked by the path validator, preventing agents from reading skill definitions. This change: - Refactors `resolve_local_tool_path` into `validate_local_tool_path`, a pure security gate that no longer resolves paths (left to the sandbox) - Permits read-only access to the skills container path (/mnt/skills by default, configurable via config.skills.container_path) - Blocks write access to skills paths (PermissionError) - Allows /mnt/skills in bash command path validation - Adds `LocalSandbox.update_path_mappings` and injects per-thread user-data mappings into the sandbox so all virtual-path resolution is handled uniformly by the sandbox layer - Covers all new behaviour with tests Fixes #1177 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(sandbox): unify all virtual path resolution in tools.py Move skills path resolution from LocalSandbox into tools.py so that all virtual-to-host path translation (user-data and skills) lives in one layer. LocalSandbox becomes a pure execution layer that receives only real host paths — no more path_mappings, _resolve_path, or reverse resolve logic. This addresses architecture feedback that path resolution was split across two layers (tools.py for user-data, LocalSandbox for skills), making the flow hard to follow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(sandbox): address Copilot review — cache-on-success and error path masking - Replace @lru_cache with manual cache-on-success for _get_skills_container_path and _get_skills_host_path so transient failures at startup don't permanently disable skills access. - Add _sanitize_error() helper that masks host filesystem paths in error messages via mask_local_paths_in_output before returning them to the agent. - Apply _sanitize_error() to all catch-all (Exception/OSError) handlers in sandbox tool functions to prevent host path leakage in error output. - Remove unused lru_cache import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-17 21:44:36 +08:00
def test_validate_local_bash_command_paths_blocks_traversal_in_user_data() -> None:
"""Bash commands with traversal in user-data paths should be blocked."""
with pytest.raises(PermissionError, match="path traversal"):
validate_local_bash_command_paths(
"cat /mnt/user-data/workspace/../../etc/passwd",
_THREAD_DATA,
)
def test_validate_local_bash_command_paths_blocks_traversal_in_skills() -> None:
"""Bash commands with traversal in skills paths should be blocked."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
with pytest.raises(PermissionError, match="path traversal"):
validate_local_bash_command_paths(
"cat /mnt/skills/../../etc/passwd",
_THREAD_DATA,
)
# ---------- Skills path tests ----------
def test_is_skills_path_recognises_default_prefix() -> None:
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
assert _is_skills_path("/mnt/skills") is True
assert _is_skills_path("/mnt/skills/public/bootstrap/SKILL.md") is True
assert _is_skills_path("/mnt/skills-extra/foo") is False
assert _is_skills_path("/mnt/user-data/workspace") is False
def test_validate_local_tool_path_allows_skills_read_only() -> None:
"""read_file / ls should be able to access /mnt/skills paths."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
# Should not raise
validate_local_tool_path(
"/mnt/skills/public/bootstrap/SKILL.md",
_THREAD_DATA,
read_only=True,
)
def test_validate_local_tool_path_blocks_skills_write() -> None:
"""write_file / str_replace must NOT write to skills paths."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
with pytest.raises(PermissionError, match="Write access to skills path is not allowed"):
validate_local_tool_path(
"/mnt/skills/public/bootstrap/SKILL.md",
_THREAD_DATA,
read_only=False,
)
def test_validate_local_bash_command_paths_allows_skills_path() -> None:
"""bash commands referencing /mnt/skills should be allowed."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
validate_local_bash_command_paths(
"cat /mnt/skills/public/bootstrap/SKILL.md",
_THREAD_DATA,
)
def test_validate_local_bash_command_paths_still_blocks_other_paths() -> None:
"""Paths outside virtual and system prefixes must still be blocked."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/mnt/skills"):
with pytest.raises(PermissionError, match="Unsafe absolute paths"):
validate_local_bash_command_paths("cat /etc/shadow", _THREAD_DATA)
def test_validate_local_tool_path_skills_custom_container_path() -> None:
"""Skills with a custom container_path in config should also work."""
with patch("deerflow.sandbox.tools._get_skills_container_path", return_value="/custom/skills"):
# Should not raise
validate_local_tool_path(
"/custom/skills/public/my-skill/SKILL.md",
_THREAD_DATA,
read_only=True,
)
# The default /mnt/skills should not match since container path is /custom/skills
with pytest.raises(PermissionError, match="Only paths under"):
validate_local_tool_path(
"/mnt/skills/public/bootstrap/SKILL.md",
_THREAD_DATA,
read_only=True,
)
feat(harness): integration ACP agent tool (#1344) * refactor: extract shared utils to break harness→app cross-layer imports Move _validate_skill_frontmatter to src/skills/validation.py and CONVERTIBLE_EXTENSIONS + convert_file_to_markdown to src/utils/file_conversion.py. This eliminates the two reverse dependencies from client.py (harness layer) into gateway/routers/ (app layer), preparing for the harness/app package split. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor: split backend/src into harness (deerflow.*) and app (app.*) Physically split the monolithic backend/src/ package into two layers: - **Harness** (`packages/harness/deerflow/`): publishable agent framework package with import prefix `deerflow.*`. Contains agents, sandbox, tools, models, MCP, skills, config, and all core infrastructure. - **App** (`app/`): unpublished application code with import prefix `app.*`. Contains gateway (FastAPI REST API) and channels (IM integrations). Key changes: - Move 13 harness modules to packages/harness/deerflow/ via git mv - Move gateway + channels to app/ via git mv - Rename all imports: src.* → deerflow.* (harness) / app.* (app layer) - Set up uv workspace with deerflow-harness as workspace member - Update langgraph.json, config.example.yaml, all scripts, Docker files - Add build-system (hatchling) to harness pyproject.toml - Add PYTHONPATH=. to gateway startup commands for app.* resolution - Update ruff.toml with known-first-party for import sorting - Update all documentation to reflect new directory structure Boundary rule enforced: harness code never imports from app. All 429 tests pass. Lint clean. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: add harness→app boundary check test and update docs Add test_harness_boundary.py that scans all Python files in packages/harness/deerflow/ and fails if any `from app.*` or `import app.*` statement is found. This enforces the architectural rule that the harness layer never depends on the app layer. Update CLAUDE.md to document the harness/app split architecture, import conventions, and the boundary enforcement test. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: add config versioning with auto-upgrade on startup When config.example.yaml schema changes, developers' local config.yaml files can silently become outdated. This adds a config_version field and auto-upgrade mechanism so breaking changes (like src.* → deerflow.* renames) are applied automatically before services start. - Add config_version: 1 to config.example.yaml - Add startup version check warning in AppConfig.from_file() - Add scripts/config-upgrade.sh with migration registry for value replacements - Add `make config-upgrade` target - Auto-run config-upgrade in serve.sh and start-daemon.sh before starting services - Add config error hints in service failure messages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix comments * fix: update src.* import in test_sandbox_tools_security to deerflow.* Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: handle empty config and search parent dirs for config.example.yaml Address Copilot review comments on PR #1131: - Guard against yaml.safe_load() returning None for empty config files - Search parent directories for config.example.yaml instead of only looking next to config.yaml, fixing detection in common setups Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: correct skills root path depth and config_version type coercion - loader.py: fix get_skills_root_path() to use 5 parent levels (was 3) after harness split, file lives at packages/harness/deerflow/skills/ so parent×3 resolved to backend/packages/harness/ instead of backend/ - app_config.py: coerce config_version to int() before comparison in _check_config_version() to prevent TypeError when YAML stores value as string (e.g. config_version: "1") - tests: add regression tests for both fixes Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: update test imports from src.* to deerflow.*/app.* after harness refactor Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(harness): add tool-first ACP agent invocation (#37) * feat(harness): add tool-first ACP agent invocation * build(harness): make ACP dependency required * fix(harness): address ACP review feedback * feat(harness): decouple ACP agent workspace from thread data ACP agents (codex, claude-code) previously used per-thread workspace directories, causing path resolution complexity and coupling task execution to DeerFlow's internal thread data layout. This change: - Replace _resolve_cwd() with a fixed _get_work_dir() that always uses {base_dir}/acp-workspace/, eliminating virtual path translation and thread_id lookups - Introduce /mnt/acp-workspace virtual path for lead agent read-only access to ACP agent output files (same pattern as /mnt/skills) - Add security guards: read-only validation, path traversal prevention, command path allowlisting, and output masking for acp-workspace - Update system prompt and tool description to guide LLM: send self-contained tasks to ACP agents, copy results via /mnt/acp-workspace - Add 11 new security tests for ACP workspace path handling Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * refactor(prompt): inject ACP section only when ACP agents are configured The ACP agent guidance in the system prompt is now conditionally built by _build_acp_section(), which checks get_acp_agents() and returns an empty string when no ACP agents are configured. This avoids polluting the prompt with irrelevant instructions for users who don't use ACP. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix lint * fix(harness): address Copilot review comments on sandbox path handling and ACP tool - local_sandbox: fix path-segment boundary bug in _resolve_path (== or startswith +"/") and add lookahead in _resolve_paths_in_command regex to prevent /mnt/skills matching inside /mnt/skills-extra - local_sandbox_provider: replace print() with logger.warning(..., exc_info=True) - invoke_acp_agent_tool: guard getattr(option, "optionId") with None default + continue; move full prompt from INFO to DEBUG level (truncated to 200 chars) - sandbox/tools: fix _get_acp_workspace_host_path docstring to match implementation; remove misleading "read-only" language from validate_local_bash_command_paths Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(acp): thread-isolated workspaces, permission guardrail, and ContextVar registry P1.1 – ACP workspace thread isolation - Add `Paths.acp_workspace_dir(thread_id)` for per-thread paths - `_get_work_dir(thread_id)` in invoke_acp_agent_tool now uses `{base_dir}/threads/{thread_id}/acp-workspace/`; falls back to global workspace when thread_id is absent or invalid - `_invoke` extracts thread_id from `RunnableConfig` via `Annotated[RunnableConfig, InjectedToolArg]` - `sandbox/tools.py`: `_get_acp_workspace_host_path(thread_id)`, `_resolve_acp_workspace_path(path, thread_id)`, and all callers (`replace_virtual_paths_in_command`, `mask_local_paths_in_output`, `ls_tool`, `read_file_tool`) now resolve ACP paths per-thread P1.2 – ACP permission guardrail - New `auto_approve_permissions: bool = False` field in `ACPAgentConfig` - `_build_permission_response(options, *, auto_approve: bool)` now defaults to deny; only approves when `auto_approve=True` - Document field in `config.example.yaml` P2 – Deferred tool registry race condition - Replace module-level `_registry` global with `contextvars.ContextVar` - Each asyncio request context gets its own registry; worker threads inherit the context automatically via `loop.run_in_executor` - Expose `get_deferred_registry` / `set_deferred_registry` / `reset_deferred_registry` helpers Tests: 831 pass (57 for affected modules, 3 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(sandbox): mount /mnt/acp-workspace in docker sandbox container The AioSandboxProvider was not mounting the ACP workspace into the sandbox container, so /mnt/acp-workspace was inaccessible when the lead agent tried to read ACP results in docker mode. Changes: - `ensure_thread_dirs`: also create `acp-workspace/` (chmod 0o777) so the directory exists before the sandbox container starts — required for Docker volume mounts - `_get_thread_mounts`: add read-only `/mnt/acp-workspace` mount using the per-thread host path (`host_paths.acp_workspace_dir(thread_id)`) - Update stale CLAUDE.md description (was "fixed global workspace") Tests: `test_aio_sandbox_provider.py` (4 new tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(lint): remove unused imports in test_aio_sandbox_provider Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix config --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-26 14:20:18 +08:00
# ---------- ACP workspace path tests ----------
def test_is_acp_workspace_path_recognises_prefix() -> None:
assert _is_acp_workspace_path("/mnt/acp-workspace") is True
assert _is_acp_workspace_path("/mnt/acp-workspace/hello.py") is True
assert _is_acp_workspace_path("/mnt/acp-workspace-extra/foo") is False
assert _is_acp_workspace_path("/mnt/user-data/workspace") is False
def test_validate_local_tool_path_allows_acp_workspace_read_only() -> None:
"""read_file / ls should be able to access /mnt/acp-workspace paths."""
validate_local_tool_path(
"/mnt/acp-workspace/hello_world.py",
_THREAD_DATA,
read_only=True,
)
def test_validate_local_tool_path_blocks_acp_workspace_write() -> None:
"""write_file / str_replace must NOT write to ACP workspace paths."""
with pytest.raises(PermissionError, match="Write access to ACP workspace is not allowed"):
validate_local_tool_path(
"/mnt/acp-workspace/hello_world.py",
_THREAD_DATA,
read_only=False,
)
def test_validate_local_bash_command_paths_allows_acp_workspace() -> None:
"""bash commands referencing /mnt/acp-workspace should be allowed."""
validate_local_bash_command_paths(
"cp /mnt/acp-workspace/hello_world.py /mnt/user-data/outputs/hello_world.py",
_THREAD_DATA,
)
def test_validate_local_bash_command_paths_blocks_traversal_in_acp_workspace() -> None:
"""Bash commands with traversal in ACP workspace paths should be blocked."""
with pytest.raises(PermissionError, match="path traversal"):
validate_local_bash_command_paths(
"cat /mnt/acp-workspace/../../etc/passwd",
_THREAD_DATA,
)
def test_resolve_acp_workspace_path_resolves_correctly(tmp_path: Path) -> None:
"""ACP workspace virtual path should resolve to host path."""
acp_dir = tmp_path / "acp-workspace"
acp_dir.mkdir()
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=str(acp_dir)):
resolved = _resolve_acp_workspace_path("/mnt/acp-workspace/hello.py")
assert resolved == str(acp_dir / "hello.py")
def test_resolve_acp_workspace_path_resolves_root(tmp_path: Path) -> None:
"""ACP workspace root should resolve to host directory."""
acp_dir = tmp_path / "acp-workspace"
acp_dir.mkdir()
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=str(acp_dir)):
resolved = _resolve_acp_workspace_path("/mnt/acp-workspace")
assert resolved == str(acp_dir)
def test_resolve_acp_workspace_path_raises_when_not_available() -> None:
"""Should raise FileNotFoundError when ACP workspace does not exist."""
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=None):
with pytest.raises(FileNotFoundError, match="ACP workspace directory not available"):
_resolve_acp_workspace_path("/mnt/acp-workspace/hello.py")
def test_resolve_acp_workspace_path_blocks_traversal(tmp_path: Path) -> None:
"""Path traversal in ACP workspace paths must be rejected."""
acp_dir = tmp_path / "acp-workspace"
acp_dir.mkdir()
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=str(acp_dir)):
with pytest.raises(PermissionError, match="path traversal"):
_resolve_acp_workspace_path("/mnt/acp-workspace/../../etc/passwd")
def test_replace_virtual_paths_in_command_replaces_acp_workspace() -> None:
"""ACP workspace virtual paths in commands should be resolved to host paths."""
acp_host = "/home/user/.deer-flow/acp-workspace"
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=acp_host):
cmd = "cp /mnt/acp-workspace/hello.py /mnt/user-data/outputs/hello.py"
result = replace_virtual_paths_in_command(cmd, _THREAD_DATA)
assert "/mnt/acp-workspace" not in result
assert f"{acp_host}/hello.py" in result
assert "/tmp/deer-flow/threads/t1/user-data/outputs/hello.py" in result
def test_mask_local_paths_in_output_hides_acp_workspace_host_paths() -> None:
"""ACP workspace host paths in bash output should be masked to virtual paths."""
acp_host = "/home/user/.deer-flow/acp-workspace"
with patch("deerflow.sandbox.tools._get_acp_workspace_host_path", return_value=acp_host):
output = f"Copied: {acp_host}/hello.py"
masked = mask_local_paths_in_output(output, _THREAD_DATA)
assert acp_host not in masked
assert "/mnt/acp-workspace/hello.py" in masked