arkiv-best-practices

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill instructs the agent to fetch and parse data from the public Arkiv Braga testnet (e.g., SKILL.md and references show publicClient.buildQuery().fetch(), publicClient.getEntity(), and JSON-RPC curl calls to https://braga.hoodi.arkiv.network/rpc), which are open/public, user-generated entities that the agent is expected to read and act on, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill includes browser CDN imports that fetch and execute remote JavaScript at runtime (e.g., https://esm.sh/@arkiv-network/sdk@0.6.0?target=es2022&bundle-deps), which is a runtime external dependency that executes remote code.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly for an Ethereum-based SDK and includes wallet-backed write operations. It documents createWalletClient/privateKeyToAccount, WalletClient methods that create/update/delete entities and return txHash, guidance about private keys, gas/insufficient funds, and using testnet faucets. These are specific crypto/blockchain wallet and transaction-signing APIs (i.e., ability to sign/send on-chain transactions), which constitute direct financial execution capability.