skills/bankrbot/skills/staq/Gen Agent Trust Hub

staq

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes curl for JSON-RPC interaction with the Base blockchain and for communicating with the STAQ API service endpoints.
  • [EXTERNAL_DOWNLOADS]: The skill retrieves transaction quotes, session nonces, and account status summaries from the external domain api.agentstaq.xyz.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from API and blockchain responses, which represents an attack surface, but it applies comprehensive defenses.
  • Ingestion points: Data sourced from https://api.agentstaq.xyz and the Base blockchain RPC via instructions in SKILL.md and supplemental reference files.
  • Boundary markers: Instructions explicitly mandate the agent to treat external text solely as data and ignore any prose-based instructions to prevent hijacking.
  • Capability inventory: Includes persistent file storage access, USDC wallet transfers, and arbitrary contract call capabilities via the Bankr platform.
  • Sanitization: The skill enforces mandatory on-chain derivation of contract addresses from a pinned hub, manual reconstruction and byte-for-byte comparison of calldata, and simulation of transactions using eth_call to verify results before execution.
  • [SAFE]: The skill adopts a trust-anchor architecture where critical addresses (Hub, Vault, USDC) are hardcoded, preventing them from being spoofed by external inputs or API responses. It requires explicit user consent for rule changes and ensures all financial operations are independently validated against immutable blockchain state.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 06:29 PM