The Agent Skills Directory

[SAFE]: The skill operates entirely within the context of text analysis and documentation review. It does not include scripts, executables, or tools that perform network operations, file system modifications, or command-line execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external content such as API design documents and OpenAPI specifications. This constitutes an attack surface where malicious instructions could be embedded in the data being reviewed. However, the skill includes robust mitigation instructions: it requires a formal audit of inputs (classifying them as known, missing, conflicting, etc. in prompts/api-design-quality-review.md), strictly prohibits executing APIs or adopting external claims as facts, and limits its own output to analytical markdown findings. There are no high-risk capabilities present that could be exploited by an indirect injection.
[INDIRECT_PROMPT_INJECTION]: Mandatory Evidence Chain: (1) Ingestion points: API designs and OpenAPI material provided in the prompt as defined in prompts/api-design-quality-review.md; (2) Boundary markers: Workflow requires classification of input as known, missing, conflicting, stale, out_of_scope, and assumptions; (3) Capability inventory: No subprocess calls, file-writes, or network operations detected across all scripts; (4) Sanitization: Absent, but restricted output format (Markdown) and analytical scope reduce impact.

api-design-quality-review