The Agent Skills Directory

Command Execution: The linter phase of the skill executes several Python scripts and shell commands to perform mechanical validations on the repository's files. These checks include verifying JSON/YAML syntax, ensuring anchor uniqueness, and resolving internal Markdown links. These operations are scoped to the local workspace and are used to enforce specification standards.
External Downloads: The skill utilizes npx to fetch and execute the @mermaid-js/mermaid-cli utility for the purpose of validating diagrams within the specification files. This involves downloading the tool from the official NPM registry at runtime, which is a standard method for accessing specialized development utilities.
Indirect Prompt Injection Surface: Because the skill analyzes the content of pull request diffs and the specification artifacts themselves, it ingests untrusted data that could potentially contain instructions designed to influence the AI agents' review outcomes. The skill manages this risk by positioning the automated findings as advisory only, ensuring that a human maintainer performs the final review and decision-making.

apm-spec-guardian