The Agent Skills Directory

[Command Execution] (SAFE): Shell scripts are used for local workspace automation, such as analyzing git history and configuring hooks. These actions are standard for development tools and do not involve remote execution.\n- [Indirect Prompt Injection] (LOW): The skill processes data from git commit messages and tool sessions. This data is then formatted into markdown reports, creating a surface for indirect prompt injection.\n
Ingestion points: Git logs in scripts/analyze-history.sh and session logs in scripts/analyze-sessions.sh.\n
Boundary markers: No explicit delimiters or warnings are used in the generated reports to distinguish between system analysis and ingested data.\n
Capability inventory: The skill can write files and execute local git/bash commands.\n
Sanitization: Uses jq for structured extraction, but does not sanitize the text content of commits or logs.

claude-reflect