The Agent Skills Directory

[SAFE]: The skill defines a disciplined debugging process for identifying and fixing software defects. No malicious patterns, obfuscation, or exfiltration attempts were detected.
[COMMAND_EXECUTION]: The skill instructions guide the agent to create and run reproduction scripts, such as the provided scripts/hitl-loop.template.sh, and use development tools like curl, git, and headless browsers. These actions are functionally necessary for software debugging and are performed within the local development environment.
[PROMPT_INJECTION]: The skill uses natural instructional language to maintain the agent's focus on structured debugging phases. It does not attempt to override safety filters or bypass core agent instructions.

diagnosing-bugs