The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill extensively documents the use of the run_script tool to execute Python code generated or provided at runtime. While it encourages short, deterministic scripts, this mechanism allows for the dynamic modification of the FreeCAD environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external websites such as componentsearchengine.com and GitHub for component research and troubleshooting.
Ingestion points: External research sources including the FreeCAD forum, GitHub Issues/Discussions, and specialized electronic component search engines (references/community-support.md, references/electronic-components.md).
Boundary markers: The instructions recommend redacting credentials and private data before drafting reports based on external findings, although it lacks technical delimiters for the ingested content itself.
Capability inventory: The skill has access to tools for file system modification (save_document, export), process execution (run_fem), and arbitrary script execution (run_script).
Sanitization: The agent is instructed to manually redact sensitive information from draft reports, but no automated sanitization of external data is defined.
[REMOTE_CODE_EXECUTION]: The run_script tool provides a documented interface for arbitrary Python execution. The skill explicitly notes that this tool is not sandboxed and operates with the user's privileges, cautioning against the inclusion of untrusted code.
[COMMAND_EXECUTION]: The skill orchestrates external processes for manufacturing tasks and engineering simulations, specifically the CalculiX solver via the run_fem tool.
[CREDENTIALS_UNSAFE]: The documentation discusses the protocol's use of bearer tokens for network access. It includes a specific security rule forbidding the logging or printing of these tokens in any output or script.

freecad-mcp