claude-verifier

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt asks the user for an "API Key" and includes example curl commands and a Python script that embed that key directly into Authorization headers (e.g., "Authorization: Bearer {api_key}"), which instructs the agent to accept and place secret values verbatim into requests/outputs.

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

• Third-party content exposure detected (high risk: 1.00). Yes — SKILL.md's required detection workflow (the Python script in "知识库截止时间判断法" and the get_user_input flow) explicitly asks for a user-provided API URL and then POSTs to that endpoint and parses the returned "content" to decide the model version, and the guide also mandates "联网搜索测试"/MCP and web_search examples that fetch arbitrary webpages; these steps ingest untrusted third‑party content and use it to drive decisions.