The Agent Skills Directory

[COMMAND_EXECUTION]: The skill installs and executes a local Python script (quality-gate.py) as a Stop hook. This script is used to mechanically enforce session hygiene by checking machine-verifiable facts like file timestamps and disk usage.
[DATA_EXPOSURE]: The script reads the session transcript and file system metadata (modification times) within the agent's internal project directory (~/.claude/projects/). This is used to verify that the model has captured learnings as instructed. No data is sent to external servers.
[SAFE]: The skill does not use any external dependencies, perform network requests, or include obfuscated code. It uses deterministic regex patterns for heuristic quality checks and follows standard practices for extending agent functionality via hooks.

delivery-gate