The Agent Skills Directory

[SAFE]: The skill consists of instructional content that outlines a structured debugging process. No code execution, network access, or persistence mechanisms are present.
[DATA_EXPOSURE_AND_EXFILTRATION]: The debugging guidelines specifically warn against the extraction of sensitive data such as credentials, tokens, and full configuration files, instructing the agent to log only non-sensitive metadata (lengths, names, or shapes).
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest external data including command outputs, logs, and stack traces in Phase 1 and 2. While explicit boundary markers are not defined in the prompt template, the skill includes sanitization guidelines to prevent the exposure of credentials or sensitive environment variables during the gathering of evidence.

systematic-debugging