dogfood

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs the agent to fill form fields with "{EMAIL}" and "{PASSWORD}" via agent-browser commands, requiring the LLM to place user-provided credentials verbatim into generated commands (high exfiltration risk).

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md directs the agent-browser workflow to open and snapshot the user-supplied Target URL, so arbitrary outsider web page text can be read at runtime and included in the agent/LLM context (indirect prompt-injection surface).