research

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). Insecure because the prompt explicitly tells the agent to "include all context from your conversation in the task" that is passed verbatim to a spawned sub-agent (and saved/sent), so any API keys, tokens, or passwords present in the conversation would be forwarded and could appear verbatim in generated files or wake messages, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.85). The required workflow spawns a sub-agent that runs gemini --yolo "[RESEARCH PROMPT]" to perform web-style research and produce a long report with citations, which typically ingests outsider-authored free text from public sources at runtime into the LLM context.