The Agent Skills Directory

[PROMPT_INJECTION]: The skill aggregates and summarizes logs from sources like kb/learnings/ and maintenance/, creating a surface for indirect prompt injection. Malicious instructions or data within these logs could potentially influence the agent's summary output.
Ingestion points: Accesses files in kb/learnings/, maintenance/, and Claude Code session JSONL logs.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the processing instructions.
Capability inventory: The skill uses Read, Grep, and Glob tools and executes local scripts via python3.
Sanitization: The skill performs format normalization but lacks explicit sanitization or filtering of instruction-like content within the logs.
[COMMAND_EXECUTION]: The skill utilizes shell commands to execute local Python scripts and references installation hooks.
Evidence: Executes python3 scripts/session_token_stats.py to calculate token usage and compare results against a baseline.
Evidence: References the use of app/hooks/ai-toolkit-statusline.sh for rendering environment status updates.

briefing