tapestry

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 extracts content from arbitrary public URLs (YouTube via yt-dlp, web articles via article-extractor/trafilatura/curl, and PDFs via curl/pdftotext) and then reads and uses that content to create action plans, exposing the agent to untrusted, user-provided third‑party content that could carry indirect prompt injections.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches arbitrary, user-supplied URLs at runtime (the $URL variable — e.g., https://www.youtube.com/watch?v=dQw4w9WgXcQ, https://example.com/how-to-build-saas, https://example.com/research-paper.pdf) via curl/yt-dlp/trafilatura and injects the retrieved content into the workflow to generate action plans, which means remote content can directly control the agent's prompts/outputs.