The Agent Skills Directory

[COMMAND_EXECUTION]: The script generate_insights.py invokes analyze_patterns.py using the subprocess module to collect usage data.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes historical chat logs which could contain malicious instructions.
Ingestion points: Chat history is read from ~/.claude/history.jsonl and ~/.claude/projects/ by the archival and analysis scripts.
Boundary markers: The prompts in references/deep_analysis_agents.md lack explicit delimiters to separate instructions from the ingested chat data.
Capability inventory: Analysis sub-agents are granted access to file system tools including Read, Glob, and Grep.
Sanitization: Historical chat content is processed without sanitization or filtering before being passed to sub-agents.

cc-insights