zhin-adapter-development

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOW
Full Analysis
  • [Category 1: Prompt Injection] (SAFE): No instructions designed to override system prompts, bypass safety filters, or extract internal configuration were found.
  • [Category 2: Data Exposure & Exfiltration] (SAFE): Credential fields in the code examples use obvious placeholders like 'bot-token-1' and 'bot-token-here'. No patterns involving the unauthorized reading of sensitive files or exfiltration to external domains were detected.
  • [Category 4: Unverifiable Dependencies] (SAFE): The imports for '@zhin.js/core' and 'zhin.js' are standard for the Zhin framework identified in the metadata. No suspicious third-party package installations or remote script executions were identified.
  • [Category 8: Indirect Prompt Injection] (SAFE): While the documentation describes how to ingest external message data, the skill itself is a static instructional resource and does not expose a live data-processing surface for the agent.
  • [Category 10: Dynamic Execution] (SAFE): The use of 'Adapter.register' and tool execution patterns are standard architectural features of the documented framework and do not involve unsafe runtime code generation from untrusted sources.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 01:06 PM