The Agent Skills Directory

[PERSISTENCE]: The skill includes configuration instructions to register a command hook within the agent's environment settings (~/.claude/settings.json). This hook ensures that the session evaluation script runs automatically every time a session terminates.
[INDIRECT_PROMPT_INJECTION]: The script processes conversation data which contains untrusted inputs from the user.
Ingestion points: The evaluate-session.sh script reads the session transcript file specified by the CLAUDE_TRANSCRIPT_PATH environment variable.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are implemented when the script parses the transcript for message counts.
Capability inventory: The script performs filesystem operations such as directory creation (mkdir) and uses grep for text matching. It does not perform network operations or execute content from the transcript.
Sanitization: The script utilizes jq to safely parse values from the config.json file.

cc-skill-continuous-learning