lavish

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

• Third-party content exposure detected (high risk: 0.85). During step 3, npx -y lavish-axi poll <html-file> long-polls for browser-reported user annotations/queued prompts, which are free-form text authored by the outsider (the human reviewer) and then ingested into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill repeatedly requires running "npx -y lavish-axi " which fetches and executes the remote npm package (e.g. from https://registry.npmjs.org/lavish-axi) at runtime and is relied upon to open/poll interactive review sessions that can influence agent prompts and execute code.