solidworks-fillet-chamfer-cnc

Warn

Audited by Gen Agent Trust Hub on Oct 10, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill dynamically compiles and executes C# and C++ code at runtime using scripts/hold_line_bridge.py to interact with SolidWorks APIs (ISetHoldLines) that are unavailable or unstable via standard Python COM bindings.
  • Evidence: scripts/native/HoldLineBridge.cs (C# source), scripts/native/NativeHoldLineAddin.cpp (C++ source), and the build_hold_line_bridge / build_native_hold_line_addin functions in scripts/hold_line_bridge.py.
  • [COMMAND_EXECUTION]: The skill executes shell commands to invoke compilers (csc.exe, cl.exe) and run the resulting compiled binaries.
  • Evidence: subprocess.run() calls in scripts/hold_line_bridge.py to trigger .NET Framework and MSVC toolchains.
  • [EXTERNAL_DOWNLOADS]: The skill downloads CAD reference files (STEP format) from the FreeCAD-library's official GitHub repository. These downloads are performed with integrity checks using SHA-256 hashes.
  • Evidence: scripts/verify_open_source_complex_case.py uses urllib.request to fetch files from raw.githubusercontent.com/FreeCAD/FreeCAD-library.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes project-specific parameters from JSON files and geometric data from imported STEP files. While it uses strict validation logic (e.g., finite_number, validate_basename), this ingestion of external data constitutes a surface for influence on the agent's tool-calling logic.
  • Ingestion points: scripts/cnc_strategy.py (JSON parameters), scripts/verify_open_source_bracket.py (STEP import).
  • Boundary markers: Uses structured JSON schema and geometric count assertions (e.g., expected_source_topology).
  • Capability inventory: File system writes, SolidWorks COM manipulation, shell command execution via subprocess.
  • Sanitization: Employs specific sanitization functions for numerical inputs and file paths.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 10, 2026, 08:20 AM