The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions mandate the use of shell commands through the git and gh (GitHub CLI) utilities. These include git merge-base, git blame, git fetch, git show, git grep, and git worktree to manage repository state, fetch remote pull request data, and analyze changed surfaces.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content originating from external pull request metadata and commit history, creating a potential surface for indirect injection.
Ingestion points: Pull request titles, bodies, and commit messages are retrieved via gh pr view and commit logs (described in SKILL.md and scope-resolution.md).
Boundary markers: The instructions lack explicit directives for the agent to use delimiters or ignore instructions that might be embedded within the PR description or code comments being reviewed.
Capability inventory: The agent has the capability to execute shell commands (git, gh) and perform network operations via git fetch to interact with remote repositories.
Sanitization: There is no mention of sanitizing or escaping the content of pull requests before the agent interprets it to form interface findings.

interface-review