The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local Python helper script (catalog.py) to manage complex metadata indexing, search, and preparation tasks. The script relies exclusively on the Python standard library and does not perform network operations or unauthorized system modifications.- [INDIRECT_PROMPT_INJECTION]: The catalog processes metadata, descriptions, and bodies from other skill files on the system. This creates a surface for indirect prompt injection where a malicious skill could include instructions in its metadata (e.g., in the description field) to influence the agent's selection process or subsequent execution phases. This risk is mitigated by the structured JSON output and the agent's internal guardrails.- [SAFE]: The Python helper script implements rigorous path validation via the safe_file function, which checks for absolute paths, parent directory traversal (..), and ensures that all resolved paths remain strictly within their designated provider roots.- [SAFE]: The script includes a _no_links check that identifies and rejects symbolic links and Windows junctions, protecting against file system redirection attacks.- [SAFE]: A 2 MiB limit is strictly enforced on all file read operations to prevent resource exhaustion or denial-of-service attacks from maliciously large files.- [SAFE]: The skill's instructions and implementation are strictly read-only and explicitly prohibit the installation of plugins, network retrieval, or the modification of file permissions.

skill-catalog