knowledge-intake

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md explicitly instructs the agent to "FETCH" user-shared external resources (web articles/blog posts via WebFetch and document URLs via document-conversion) and to read/evaluate that content as part of the intake workflow (evaluation, routing, storing, applying, and promotion to GitHub Discussions), meaning untrusted third‑party pages can be ingested and can materially influence subsequent tool use and actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches and injects external web content at runtime (e.g., WebFetch of "https://konmari.com/about-the-konmari-method/") into the agent's processing pipeline for evaluation/storage, meaning that fetched content can directly influence agent prompts and behavior.