analyze-with-file

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The workflow explicitly scans and extracts content from the codebase and writes findings into discussion.md, JSON artifacts, and CLI/echo commands (e.g., issue creation, handoff JSON) without any redaction or secret-handling safeguards, so any secrets in repo files or configs could be copied verbatim into outputs and external commands.

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly calls web.run in Phase 2 ("External Research") and Phase 3 ("外部研究" flow) to fetch public web search results which are written into research.json/explorations.json and then used to inform recommendations and next actions, meaning untrusted third‑party content from the open web is ingested and can materially influence decisions.