fetcher

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

• Third-party content exposure detected (medium risk: 0.30). The fetcher runtime workflow ingests outsider-authored free text from user-provided URL lists (e.g., via fetcher get-manifest - reading URLs from stdin), then fetches and extracts the page/PDF content into result.text/text_blobs for downstream LLM-friendly markdown.

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

• Potentially malicious external URL detected (high risk: 0.90). At runtime the wrapper invokes uvx/uv with a remote git+https://github.com/grahama1970/fetcher.git (pinned commit git+https://github.com/grahama1970/fetcher.git@35c4983729d05ca3fd7825b04c67b518f5567a6c) to fetch and execute the fetcher tool (and to run playwright install), so remote code from that URL is fetched and executed during skill runtime.