The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PR comments (human and bot) retrieved via npx agent-reviews and CI logs via gh run view. There are no specified boundary markers to separate this untrusted content from instructions, and the agent has significant capabilities including git commit and git push. No sanitization is performed on the comment content before evaluation.
[EXTERNAL_DOWNLOADS]: The skill uses package runners like npx to download and execute the latest version of the agent-reviews package from the NPM registry at runtime.
[COMMAND_EXECUTION]: The agent executes git push and git commit commands based on logic derived from the content of external PR comments, creating a risk of unauthorized modifications to the codebase.

resolve-reviews