vuln-diagnose

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly requires producing "exact HTTP request" and "HTTP request + response" evidence and even mentions exfiltrating a cookie and using a user token, which can force the model to include cookies/session tokens or other secrets verbatim in its output.

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md the required workflow asks the user/agent to construct and run attacker-chosen HTTP requests/payloads against endpoints/components (e.g., “Create the minimal test case… An HTTP request… A payload… Response… retrieve data belonging to another user”), which means it can ingest outsider-authored free text at runtime via whatever request/response evidence the user supplies to be documented.