web-doc-resolver

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

• Third-party content exposure detected (high risk: 1.00). The skill (SKILL.md and scripts/resolve.py) explicitly fetches and ingests content from arbitrary public websites and search results via webfetch/websearch and llms.txt, returning that untrusted third-party content as LLM-ready markdown that the agent reads and can use to drive subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The resolver script explicitly performs a runtime fetch of each site's llms.txt (e.g., https://{origin}/llms.txt) via the webfetch subprocess and, if present, uses that fetched text as the authoritative content returned to the agent—meaning remote llms.txt files can directly control the agent's instructions at runtime.