agent-browser

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes multiple examples that embed plaintext credentials and sensitive tokens directly into commands and form fills (e.g., fill "password123", card number, state files with plaintext session tokens), which would require an LLM to produce or relay secret values verbatim—creating a high exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly lets the agent open arbitrary public URLs (agent-browser open ), snapshot and extract page content (agent-browser snapshot -i, get text, snapshot --json) and run eval in page context, and those page-derived outputs are intended to be read and acted on by the agent by default (content-boundaries are opt-in), exposing it to untrusted third‑party content that could carry indirect prompt injection.