The Agent Skills Directory

[SAFE]: The skill uses established command-line tools (git, gh) to perform its stated functions. These operations are restricted to the local repository and the authenticated user's GitHub account.
[SAFE]: The dynamic workflow script in references/review-fix-workflow.md uses standard platform APIs for orchestrating sub-agents. It implements safety-conscious patterns, such as using isolation: 'worktree' to prevent parallel agents from interfering with the main branch and requiring adversarial verification of findings before attempting fixes.
[SAFE]: No indicators of prompt injection, data exfiltration, credential harvesting, or obfuscation were found in any of the skill files.

shipping-pr-reviews