The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions involve analyzing external artifacts like code diffs and pull requests, which serve as an ingestion point for untrusted data. * Ingestion points: The references/review.md file directs the agent to read hunks and affected callers from PRs or branches. * Boundary markers: The skill does not specify the use of delimiters or boundary markers when interpolating untrusted code into the agent's context. * Capability inventory: The skill does not define direct command execution, network access, or file-write capabilities, limiting the potential impact of an injection. * Sanitization: No sanitization or filtering of external code content is defined in the instructions.
[NO_CODE]: The skill is comprised of Markdown and YAML documentation only and does not contain any executable scripts or binaries, which significantly reduces the technical risk surface.

code-review