reference-validator

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill requires calling web_fetch and returning literal content snippets (first 2000 chars) and "quote the exact characters found", which forces the model to include any secret tokens embedded in fetched pages or URLs verbatim and therefore presents a high exfiltration risk.

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

• Third-party content exposure detected (low risk: 0.10). The skill reads outsider-authored free text only when a caller provides a JSON list containing attacker-controlled url and quote, and then it re-fetches those URLs at runtime via web_fetch(url) to search for the quoted text.