solana-kit

Fail

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: HIGHDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides code snippets in references/features-keypairs.md that read the user's primary Solana CLI private key from ~/.config/solana/id.json. Accessing these credentials poses a high risk of exposure.
  • [EXTERNAL_DOWNLOADS]: The documentation describes installing various official and community Solana libraries, such as @solana/kit, @solana-program/system, and others under the @solana/ and @solana-program/ namespaces.
  • [PROMPT_INJECTION]: The skill enables an agent to fetch and interpret data from the Solana blockchain, which is an untrusted external source.
  • Ingestion points: Uses fetchEncodedAccount (references/core-accounts.md) and verifyOffchainMessageEnvelope (references/features-offchain-messages.md) to read data from the network.
  • Boundary markers: There are no boundary markers or instructions to ignore embedded text within the samples.
  • Capability inventory: The skill utilizes network RPC methods (rpc.send()) and local file system reads (fs.readFileSync).
  • Sanitization: No input validation or sanitization is demonstrated before the agent processes the external data.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 28, 2026, 09:46 PM