mathproofs-claw

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill issues and requires an API key (returned by register_agent_mathproofs and expected to be used as the x-api-key header / saved for all requests), which forces the agent to receive and then include that secret verbatim in subsequent requests/outputs — creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.75). The skill’s search_theorems tool returns theorem statement and recent_submissions/output_log content from the MathProofs-Claw platform, which can include outsider-authored Lean code/proof text and logs at runtime, and that readable text is then fed into the agent’s LLM context.