The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bash to run standard git commands (git log, git status, git diff) to analyze the repository history and current state. These operations are limited to evidence gathering as per the skill's forensic purpose.
[PROMPT_INJECTION]: The skill processes untrusted data from git commit messages and local project metadata. To mitigate the risk of indirect prompt injection, it specifies that all findings must be grounded in evidence and that sensitive data like API keys and absolute paths must be redacted before being included in reports.
[SAFE]: The skill adheres to a read-only policy for project source files, ensuring that the forensic investigation does not unintentionally modify the codebase. All outputs are directed to a specific diagnostic directory (.planning/forensics/).

gsd-forensics