synapse
Pass
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches official NATS project binaries from binaries.nats.dev for server and CLI installation.
- [COMMAND_EXECUTION]: Utilizes sudo for moving binaries and installing systemd services for agent persistence.
- [PROMPT_INJECTION]: Demonstrates an architecture vulnerable to indirect prompt injection where agents process untrusted input from other agents.
- Ingestion points: SDK request handlers in python.md, typescript.md, and go.md.
- Boundary markers: Absent in reference implementations.
- Capability inventory: Example agents integrate with Anthropic and OpenAI LLM APIs and perform messaging operations via NATS.
- Sanitization: Inputs are interpolated directly into LLM prompts without validation or escaping in the code samples.
Audit Metadata