The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses python3 to execute a local analysis script session_summary.py that processes internal session data.
[PROMPT_INJECTION]: The skill parses conversation history from local .jsonl session logs, which creates a potential surface for indirect prompt injection if those logs contain adversarial instructions.
Ingestion points: Conversation messages are read from JSONL files located in ~/.claude/projects/ by the scripts/session_summary.py script.
Boundary markers: The script output for the 'Last message' summary does not employ specific delimiters or instruction isolators to protect the agent's context from embedded commands.
Capability inventory: The script itself is limited to read-only file operations, although the agent platform executing the skill typically possesses shell access.
Sanitization: Extracted log text is truncated to 200 characters for display but does not undergo specific sanitization for prompt injection patterns.

claude-context