The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from git history and diffs to identify lessons learned. Security is maintained through a strict human-in-the-loop requirement where the agent must propose changes for review rather than performing automated writes, mitigating potential injection risks from codebase content.- [COMMAND_EXECUTION]: The instructions utilize standard, read-only git commands (diff, status, log) to analyze the repository state. These operations are limited to local context and do not involve untrusted input execution.- [CREDENTIALS_UNSAFE]: The skill mentions sensitive environment variables like Stripe secrets in the context of providing examples of good documentation practices. It correctly advises that such secrets should be kept in local environment files, aligning with security best practices.

learnings