create-output-format

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow’s Step 1 (“Research”) uses WebFetch to fetch and read user-provided documentation links at runtime; if those links point to public web pages (outsider-authored free text), that fetched prose is ingested into the agent’s LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent to "Fetch and read all provided documentation using WebFetch", meaning user‑provided documentation URLs (fetched at runtime via WebFetch) will be retrieved and used to control how the agent composes the output files, so these runtime external URLs are required and directly influence agent prompts and behavior.