grounded-research

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill mandates extracting word-for-word quotes from user-provided documents and outputting them verbatim, so if those sources contain API keys, tokens, or passwords the model is required to reproduce them, creating a direct exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly requires ingesting and extracting word-for-word quotes from "documents, web pages, or long-form sources" and to record sources with access methods like "web fetch" or "search" (see SKILL.md "Extract Before Analysing" and the "Sources Consulted" response-format rules), so the agent will read and act on untrusted public web content as part of its mandatory workflow, enabling indirect prompt injection risk.