local-testing

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

• Third-party content exposure detected (high risk: 0.90). This skill's SKILL.md explicitly instructs the agent to open arbitrary web URLs (e.g., "agent-browser open https://example.com/form" in the Core Workflow) and to snapshot/eval DOM content and JS (snapshot -i, eval, get text) which the agent parses and acts on (click/fill/press), meaning untrusted third‑party page content can influence subsequent tool actions.