The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands (gh, git, ax) to interact with GitHub repositories. Security is maintained through a mandatory 'preflight' authentication check and a strict requirement for the agent to display commands and obtain explicit user consent before any account-mutating action. The use of single-quoted heredocs ('EOF') in script templates prevents shell injection from dynamic content like error logs.
[DATA_EXFILTRATION]: The skill transmits information to GitHub when filing issues. It addresses potential data exposure by instructing the agent to scrub paths, tokens, and secrets from the error output before submission. The process is further secured by a mandatory user review and confirmation of the final report content prior to execution.

ax-repo