prove

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Research phase explicitly instructs using external tools (Nia/Perplexity MCPs, "WebSearch" and "WebFetch") and downloads Mathlib via lake build, meaning the agent will fetch and read public web pages/papers and third-party repository content whose instructions or content can influence proof strategy and subsequent automated actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's setup instructs running a remote install script via curl piped to sh (https://raw.githubusercontent.com/leanprover/elan/master/elan-init.sh), which fetches and executes remote code and is presented as a required dependency for the skill.