The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a utility shell script find-polluter.sh and various command-line examples in SKILL.md and root-cause-tracing.md (such as npm test, codesign, and security). These are standard diagnostic tools and examples intended for local execution during software debugging and are consistent with the skill's primary purpose.
[PROMPT_INJECTION]: The instructions use strong imperative language (e.g., "铁律" / Iron Law, "Must complete current phase") to enforce a specific engineering methodology. These are behavioral constraints designed to improve the agent's technical performance and do not attempt to bypass core AI safety filters or guidelines.
[DATA_EXFILTRATION]: Examples in SKILL.md demonstrate safe patterns for verifying the existence of environment variables and secrets (e.g., checking if a variable is set without revealing its content) during multi-layer system debugging. These practices align with secure development standards.

systematic-debugging