web-to-native

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md’s required runtime workflow includes “capture the web original” via agent-browser open <web-url>/<route>?<params> and snapshot --json/read, which ingests outsider-authored free text from a deployed URL or arbitrary route into the agent’s LLM context.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires installing and running the external CLI tools "agent-browser" and "@swmansion/argent" at runtime (e.g. npm i -g agent-browser and npm i -g @swmansion/argent), which fetch and execute remote code from the npm registry (e.g. https://www.npmjs.com/package/agent-browser and https://www.npmjs.com/package/@swmansion/argent) and are required for the verification step.