The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze error messages, stack traces, and logs from the system. This represents an attack surface where adversarial content embedded in external log data could attempt to influence the agent's diagnostic logic. The skill mitigates this through a mandatory four-phase verification process and manual tracing requirements.
[COMMAND_EXECUTION]: The skill includes a bash script (find-polluter.sh) and provides examples of shell commands for environment diagnostics (e.g., security, codesign, env). These commands are intended for local system analysis during debugging sessions to identify where data flow or configurations fail.
[DYNAMIC_EXECUTION]: The debugging process requires the agent to generate and execute failing test cases to verify the identified root cause. This involves runtime script creation and execution of local testing frameworks, which is a standard part of the software development lifecycle but involves executing code generated during the session.

systematic-debugging