even-g2-bridge

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Communicates with external services including the Anthropic API (api.anthropic.com), OpenAI API (api.openai.com), and Telegram Bot API (api.telegram.org) to provide the bridge functionality. It also fetches data from a user-configured Gateway URL.\n- [PROMPT_INJECTION]: The skill processes user-supplied voice transcription data from G2 glasses and passes it to LLM providers. As a proxy, it is susceptible to indirect prompt injection if the ingested transcription contains malicious instructions.\n
  • Ingestion points: worker.js reads user message content from the incoming POST request body.\n
  • Boundary markers: No explicit input delimiters or instructions to ignore embedded commands are used when forwarding content to the LLM Gateway or fallback API.\n
  • Capability inventory: The skill has network access via the fetch API to communicate with external Gateway, Telegram, OpenAI, and Anthropic endpoints.\n
  • Sanitization: Input data is passed to the LLMs without sanitization; output content is cleaned and truncated solely to accommodate the hardware display limitations of the G2 glasses.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:48 AM