brian-api

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a workflow that ingests natural language intents to generate and execute blockchain transactions, creating a surface for indirect prompt injection attacks. \n
  • Ingestion points: Natural language text is ingested via the prompt parameter in brian.transact calls across multiple example files (e.g., examples/swap-intent/README.md) and the BrianClient template. \n
  • Capability inventory: The skill provides templates that use walletClient.sendTransaction (viem) and wallet.sendTransaction (ethers.js) to sign and broadcast transactions, enabling significant financial actions. \n
  • Boundary markers: No specific delimiters or boundary markers are used when interpolating user prompts into the API requests. \n
  • Sanitization: The skill provides documentation on 'Safety Guards' and 'Security Considerations' suggesting transaction simulation and manual validation, but these are not enforced by the core SDK wrapper code itself.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with the Brian API (api.brianknows.org) to fetch transaction data and DeFi information, which is the intended purpose of the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 12:31 AM
Security Audit — agent-trust-hub — brian-api