The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to create and run executable 'feedback loop' artifacts, such as failing tests, CLI fixtures, and browser scripts, to confirm bug reproduction and verify fixes. Evidence for this capability is found throughout the workflow description in SKILL.md.
[EXTERNAL_DOWNLOADS]: The workflow involves using network-capable tools like curl or HTTP scripts to interact with user-specified endpoints for the purpose of capturing bug symptoms, as described in the 'Feedback loop gate' section of SKILL.md.
[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were identified. The skill includes explicit checkpoints and 'Stop' gates to prevent the agent from applying fixes without verified reproduction, aligning with security best practices for automated debugging workflows.

tk-diagnosing-bugs