delegate

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill intentionally provisions remote code execution in a reusable 7-day sandbox with an open-by-default network option, repo mounts, and persisted/stored interactions — enabling easy remote execution, data exfiltration, and persistent access if abused.

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

• Third-party content exposure detected (medium risk: 0.30). The skill ingests the user-provided task prompt from /antigravity:delegate into the Gemini managed agent (antigravity-preview-05-2026) via interaction = client.interactions.create(..., input="<prompt>", tools=[google_search,url_context] ), enabling the agent to read outsider-authored web content selected/fetched by the model.