The Agent Skills Directory

[COMMAND_EXECUTION]: Includes a utility script find-polluter.sh designed to automate the process of finding tests that create unwanted files or state by running npm test on individual files.
[COMMAND_EXECUTION]: Provides instructional examples for using system tools like codesign, security, and git to gather diagnostic information during debugging.
[PROMPT_INJECTION]: Employs strong instructional language to guide the agent toward a thorough debugging process. These instructions are focused on technical rigor and do not target safety filter bypass or behavior override beyond the scope of debugging.
[PROMPT_INJECTION]: The skill instructs the agent to analyze error messages and stack traces. This involves processing untrusted data; however, the framework emphasizes diagnostic analysis and evidence gathering over direct execution, providing a natural buffer against injection in log data.
[DATA_EXFILTRATION]: Contains examples for safely verifying the presence of secrets in the environment without exposing their values.

systematic-debugging