chat-adapter-imessage

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external iMessage senders via the startGatewayListener method.
  • Ingestion points: Incoming messages are received and mapped to the iMessageGatewayMessageData interface within SKILL.md.
  • Boundary markers: The provided documentation and code examples do not explicitly show the use of delimiters or "ignore" instructions when passing user-supplied content to the AI model.
  • Capability inventory: The adapter provides functional capabilities to postMessage, editMessage, deleteMessage, and react within iMessage threads.
  • Sanitization: While the adapter performs markdown-to-plain-text conversion for outgoing messages, there is no evidence of sanitization or filtering of incoming prompt-injectable content at the adapter level.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 05:16 PM
Security Audit — agent-trust-hub — chat-adapter-imessage