The Agent Skills Directory

[SAFE]: The skill performs read-only scanning of the local repository structure to inventory marketplace bundles, agents, and skills. All operations are confined to the local filesystem.- [COMMAND_EXECUTION]: The scan-planning-inventory.py script uses subprocess.run to invoke a sibling script (scan-marketplace-inventory.py). This execution is well-controlled, using hardcoded argument lists without shell involvement, posing no risk of command injection.- [DATA_EXPOSURE]: The skill accesses local files (markdown, Python) to extract metadata and dependency information. This data is processed locally and written to designated temporary directories within the project structure, with no network exfiltration observed.- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from marketplace files which could potentially contain malicious instructions. However, the skill treats this data as structured metadata for inventory purposes, and the risk of the agent executing embedded instructions is negligible.

tools-marketplace-inventory