The Agent Skills Directory

[SAFE]: The skill utilizes standard version control tools (git, gh) to perform its stated function of code review and PR management. It operates within the context of a local developer environment.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data (code diffs and chat history) and has the capability to write to repositories. However, it mitigates this risk through explicit workflow boundaries.
Ingestion points: The workflow in SKILL.md reads diffs, uncommitted changes, and user intent from recent chats.
Boundary markers: The Codex publication boundary section explicitly instructs the agent to treat review and publication as distinct effects and to verify authorization for each external write.
Capability inventory: The skill uses git commit, git push, and the gh CLI to interact with remote repositories.
Sanitization: The instructions include a manual review step for security and intent fit, though no automated sanitization scripts are provided.
[EXTERNAL_DOWNLOADS]: The skill metadata in MIRROR.md references a repository belonging to a well-known service (github.com/cursor/plugins.git) as its source. This reference is documented neutrally as part of the mirror's provenance.

cursor-review-and-ship