jupiter-vrfd
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a local JavaScript template in
examples/verify.mdfor agents to facilitate transaction signing and execution in a local environment. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard, well-known Node.js libraries
@solana/web3.jsandbs58to interact with the Solana blockchain and Jupiter APIs. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied token metadata and descriptions which are sent to external API endpoints.
- Ingestion points: Token IDs, descriptions, Twitter handles, and metadata objects provided by users in
SKILL.mdandexamples/verify.md. - Boundary markers: The instructions include explicit 'Agent Operating Rules' and 'Use/Do Not Use' constraints to guide agent behavior, though the metadata payloads are passed directly to the API.
- Capability inventory: Uses network requests to the vendor's API (
api.jup.ag) and local shell execution for running the provided signing scripts. - Sanitization: Normalizes social media handles and trims token mint addresses, relying on server-side validation for the metadata content.
Audit Metadata