The Agent Skills Directory

[SAFE]: The skill implements a structured debugging workflow (Phase 1-6) without any malicious instructions. It prioritizes creating reproducible feedback loops and minimizing test cases.
[SAFE]: The human-in-the-loop template (scripts/hitl-loop.template.sh) uses standard bash commands (read, printf) to facilitate communication between the agent and the user. It does not execute arbitrary code or exfiltrate sensitive data.
[SAFE]: The skill includes explicit instructions for data safety, requiring the agent to redact secrets and credentials before showing outputs or artifacts.
[SAFE]: Network operations mentioned (like curl or Playwright) are described within the context of standard developer testing practices against local or dev environments.
[SAFE]: No obfuscation, privilege escalation, or persistence mechanisms were found.

diagnosing-bugs