web-fetcher

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and ingests arbitrary public web pages (via Jina Reader, defuddle.md, markdown.new, OpenCLI, and raw HTTP) as shown in SKILL.md/README and implemented in scripts/fetch.py, so untrusted third‑party content can be read and materially influence the agent's subsequent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill makes runtime HTTP requests to external content transformers (e.g., https://r.jina.ai/, https://defuddle.md/, https://markdown.new/) and returns that fetched page text/markdown into the agent context, meaning remote content fetched at runtime can directly control prompts/inputs to the model.