The Agent Skills Directory

[SAFE]: The skill contains no executable code or external dependencies. It strictly provides a methodology for the agent to follow during debugging sessions.
[COMMAND_EXECUTION]: The skill provides example shell commands for diagnostic instrumentation (e.g., env | grep VAR). These are benign examples provided to the agent for debugging the user's local environment during an active session.
[DATA_EXPOSURE]: While the methodology suggests reviewing configuration and environment differences, it does not attempt to access or exfiltrate sensitive files (like .ssh or .aws credentials). It remains within the scope of the user's project debugging.
[PROMPT_INJECTION]: The skill uses instructional language to define agent behavior (e.g., "NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST"), which is a standard use of system instructions and does not constitute a malicious bypass attempt.

systematic-debugging