C Ecosystem

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow includes fetching C documentation via Context7 MCP from a public web library (cppreference) (e.g., get-library-docs for topics like malloc), which can ingest outsider-authored free text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill declares runtime Context7 actions (e.g., get-library-docs context7CompatibleLibraryID="/websites/cppreference_com" topic="malloc") which fetch external documentation at runtime and can be injected into the agent's guidance/prompts, so /websites/cppreference_com is a runtime external dependency that can influence prompts.