The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external debug log data which presents a potential injection surface.
Ingestion points: extract_session_metrics.py and prune_chat_sessions.py read .jsonl log files and SQLite databases from the local VS Code workspaceStorage directory.
Boundary markers: The SKILL.md instructions explicitly direct the agent to treat log content as data and avoid executing it as instructions.
Capability inventory: The skill's scripts are capable of reading, writing, and deleting files within the workspace storage and modifying local session databases.
Sanitization: The scripts use standard JSON and SQLite libraries for data extraction and aggregation, focusing on numeric metrics rather than text execution.

analyze-copilot-sessions