The Agent Skills Directory

[SAFE]: The skill defines a defensive workflow that requires isolating reproduction attempts in disposable environments or sandboxes. It establishes clear authorization boundaries to distinguish between diagnosis and production fixes, ensuring the agent does not exceed its mandate.
[COMMAND_EXECUTION]: The agent is guided to create and run specific, minimized reproduction commands such as unit tests, CLI scripts, or HTTP checks. These actions are restricted to non-production environments and temporary harnesses as per the safety guidelines.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns or external script downloads were identified. The included shell template is a simple interactive utility designed for human-in-the-loop verification within a local context.

diagnosing-bugs