raydium
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Solana blockchain and external Raydium APIs, creating an attack surface for indirect prompt injection.
- Ingestion points: The agent ingests untrusted data through token metadata (names, symbols, and URIs pointing to Arweave/IPFS) in the token launch examples in
examples/launchlab/README.md, and through search query results and pool data from the Raydium API (api-v3.raydium.io). - Boundary markers: The instructions do not define clear boundaries or provide the agent with guidance to ignore natural language commands that might be embedded within token metadata or API responses.
- Capability inventory: The skill provides a wide range of sensitive capabilities, including the ability to execute token swaps, create liquidity pools, and launch new tokens using the Raydium SDK across multiple example files.
- Sanitization: The provided code examples and instructions lack validation or sanitization logic to filter or escape instructions contained within the external data being processed.
Audit Metadata