restaurant-booking

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

• Third-party content exposure detected (high risk: 0.75). At runtime the skill uses Orthogonal/Notte to navigate to an outsider-controlled public booking site (e.g., OpenTable/Resy) and then performs a page/scrape step, which can ingest scraped page text (outsider-authored) into the agent/LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill issues runtime API calls to the Gooseworks proxy and Orth/Notte APIs (e.g., $GOOSEWORKS_API_BASE -> https://api.gooseworks.ai and https://api.orth.sh/v1/run) which execute remote browser-automation actions (remote code/instructions) and are required for the skill to run.