The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script configured as a 'Stop' hook in the agent's settings. This script performs automated, deterministic checks of system resources and file metadata to verify session hygiene.
[DATA_EXFILTRATION]: The script reads the agent's session transcript and specific project files to verify activity. All data processing is performed locally on the user's machine, and no network exfiltration patterns or external communication tools were identified.
[PROMPT_INJECTION]: The skill uses regular expressions to scan for specific phrasing in transcripts that might indicate shortcut-taking or rationalization. This is used as a defensive feedback mechanism and does not override agent instructions or safety filters.

delivery-gate