browser-pilot

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill instructs the agent to fill login forms and to "read credentials...from local .env files or prompt inputs" and to pass values to fill_form (uid + value), which implies the LLM will receive and embed secret values verbatim in tool calls/outputs, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md describes runtime browser automation (e.g., Workflow A “Navigate” to a target URL and “take_snapshot” to extract page text), which implies the LLM will ingest outsider-authored free text from arbitrary web pages loaded at runtime.