task-groundwork

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). High-risk credential exposure: TransportConfig::serverOptions() returns a sensitive service_token which is exported into a page script and callback and is also returned by ServerWorker, allowing service credentials to leak to the browser.

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow reads the “supplied task source as authority” and repository evidence from the provided workspace (e.g., evals/files/.../input/task.md / task.txt plus repo instructions), so outsider-authored free text is directly ingested into the LLM context via the task file/content path.