The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to gather context and perform reviews.
Evidence: Uses git diff, git status, git log, git show, gh pr view, gh pr diff, gh issue view, rg, and find as documented in SKILL.md.
Note: The skill also includes templates for running language-specific test runners (e.g., pytest, npm test, cargo test) but explicitly requires user consent or specific instructions to do so.
[PROMPT_INJECTION]: The skill processes untrusted external data, creating an indirect prompt injection surface.
Ingestion points: Reads local code diffs via git and remote GitHub PR/Issue content via gh (defined in SKILL.md).
Boundary markers: Absent; the instructions do not explicitly tell the agent to ignore instructions embedded within the code or comments it reviews.
Capability inventory: Includes shell command execution (git, gh, rg) and test suite execution (e.g., npm test, pytest) as documented in SKILL.md.
Sanitization: Absent; the skill processes the ingested content without explicit filtering or escaping.
Note: This surface is common for code review tools, and the risk is mitigated by the skill's strict focus on structured reporting and safety rules prohibiting autonomous file or repository modifications.

review