context-engineering

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly recommends using MCP integrations (e.g., "Context7 — Auto-fetches relevant documentation for libraries") and instructs loading "external documentation" as part of the context hierarchy (Level 3), which means the agent will fetch and read third‑party/public documentation that could contain instruction-like content and so influence actions.