skills/bankrbot/skills/juicebox-v6/Gen Agent Trust Hub

juicebox-v6

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust transaction safety pipeline in references/modules/jb-tx-safety.md. This pipeline requires the agent to follow a strict 'review → re-verify → simulate → gas → send → prove' sequence for every write operation, ensuring no transaction is signed without user review and local verification.
  • [SAFE]: It utilizes a centralized trust anchor in references/shared/deployment-manifest.json, which stores cryptographic hashes for contract bytecode, ABI integrity, and provenance metadata. This prevents the agent from interacting with unverified or malicious smart contracts.
  • [SAFE]: A strong security boundary for generated code is defined in references/frontend-security.md. It explicitly pins dependencies to exact versions (viem@2.55.19, ethers@6.15.0), forbids execution of code from CDNs, and mandates a restrictive Content Security Policy (CSP) for all generated web interfaces.
  • [SAFE]: Utility scripts such as scripts/verify-reviewed-manifest.mjs are provided for manifest maintenance and use standard Node.js modules to verify the integrity of the skill's deployment data against live blockchain states.
  • [SAFE]: Interactions with external services (Relayr, Bendystraw) are performed using pinned endpoints, and the instructions explicitly state that data from these services must be treated as untrusted and verified locally before use.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 07:45 AM