journal-review

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill requires the model to "preserve all vulnerability fields" and emit a corrected journal, which can force the LLM to echo any secrets (API keys / tokens / ACCESS_TOKEN values) present in the input JSON verbatim, creating an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow reads outsider-authored free text from the user-provided input journal JSON (e.g., input.json passed on the required command line), which is then parsed and used to derive the reviewed severities/categories (load_json() → review_vulnerability() consuming type/description/file/line fields).