The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell commands including git diff, git fetch, git grep, and git worktree to analyze repository changes. These commands are used to resolve review targets and identify surfaces affected by code modifications without altering the user's active working tree.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from pull request titles, bodies, and commit messages to verify if the interface changes align with the author's stated intent. This introduces a surface for indirect instructions hidden within the PR metadata.
Ingestion points: Pull request metadata (title and body) retrieved via gh pr view and commit subjects.
Boundary markers: The skill does not employ specific delimiters to isolate the PR content from its internal review instructions.
Capability inventory: The skill has the capability to read any file in the repository's history and perform network fetches from the remote origin.
Sanitization: The instructions do not outline any sanitization or validation of the text retrieved from the pull request before it is used for evaluation.

interface-review