The Agent Skills Directory

Command Execution: The skill uses shell commands like git, gh, and make to manage repository state, fetch pull requests, and run linters. These operations are necessary for a documentation review workflow but require the agent to handle branch names and PR identifiers provided in user arguments.
Indirect Prompt Injection: The skill ingests untrusted data by reading pull request diffs and file content from the repository. If these files contain malicious instructions disguised as documentation, they could potentially influence the agent's behavior, although the skill's specific instructions to report against a style guide help mitigate this risk.
External Content Retrieval: The make lint_prose command is configured to automatically download and install the Vale linter into the .bin/ directory if it is missing. This is a standard development practice for ensuring consistent tooling across environments, but it involves fetching an external binary at runtime.

docs-review