The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a utility script find-polluter.sh that automates running npm test on multiple files to identify state-polluting tests. It also provides examples of using local tools like security and codesign for diagnosing build and signing issues.\n- [INDIRECT_PROMPT_INJECTION]: The skill instructions involve processing potentially untrusted data such as log files and error messages.\n
Ingestion points: The skill analyzes captured test logs (e.g., /tmp/debug-run.log) and error messages from the environment as described in root-cause-tracing.md.\n
Boundary markers: The skill explicitly mandates the redaction of secrets in any displayed or sent output in SKILL.md.\n
Capability inventory: The skill has the capability to execute shell commands and test suites for verification as seen in find-polluter.sh and SKILL.md.\n
Sanitization: There is a core rule to redact secrets with <REDACTED> and ensure credentials remain in the environment rather than in outputs.

systematic-debugging