The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/check_pack.py to validate the structure, design, and security of skill packs under development. The script relies solely on the Python 3 standard library and performs static analysis without executing the code it scans.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-authored files, including JSON metadata cards and skill pack directories, creating a surface where malicious content could attempt to influence the agent's behavior during the review process.
Ingestion points: SKILL.md (Checklist steps 1-3) instructs the agent to read and process content from knowledge.json, design.json, and the files within a target skill pack directory.
Boundary markers: The agent's instructions do not specify explicit delimiters or "ignore instructions" prompts when reading the content of the pack being validated.
Capability inventory: The skill possesses file read access and the ability to execute its internal validator script via python3.
Sanitization: The check_pack.py script identifies hardcoded secrets and structural errors but does not sanitize content to prevent prompt injection against the agent itself.

build-pack