solana-agent-kit

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). Most URLs are standard blockchain/docs endpoints (Solana RPC, arweave metadata, GitHub repo, official docs), but the presence of a raw GitHub install script (raw.githubusercontent.com/.../install.sh) and instructions to curl/execute or npx/run packages (solana-mcp) makes this a risky download/execute pattern unless you verify the repository and inspect the script first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime API call agent.methods.executeBlink({ blinkUrl: "https://example.com/blink", ... }), which clearly passes a remote blinkUrl that the agent will fetch and execute at runtime (allowing external content to control actions/execution), so this external URL pattern is a high-confidence runtime dependency.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill is explicitly a Solana blockchain agent toolkit with built-in, named actions for moving and managing funds: deployToken, transfer (SOL/SPL), trade/swaps via Jupiter, stake, bridge (Wormhole), limitOrder, perpetualTrade, lend, airdrop/sendCompressedAirdrop, create liquidity pools, mint/list NFTs, and MCP tools like TRANSFER and TRADE. It requires a private key (SOLANA_PRIVATE_KEY / KeypairWallet) and supports autonomous execution (agent.analyze + agent.execute) allowing the agent to perform on-chain transactions and rebalance portfolios automatically. These are specific crypto/financial execution capabilities (wallet signing, swaps, transfers, bridging, market orders), not generic tooling, so this grants direct financial execution authority.