skill-forge

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md Phase 1A "Research (URL-based)" workflow explicitly instructs the agent to crawl and ingest arbitrary README.md, /docs/, API references, guides, and other pages from user-supplied URLs (and to generate sources.yaml), meaning the agent will read and act on untrusted public third‑party content that can materially influence subsequent tool use and outputs.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly performs runtime crawls of external docs (notably checking for an AI-readable doc map like https://resend.com/llms.txt or /llms.txt) and uses the fetched content to build SKILL.md and drive the agent's responses, so remote content can directly control prompts/instructions and is a required dependency for URL-based forging.