Execution Agent

Module Role

The Execution Agent is a specialized component within the Aarna agentic architecture. It receives validated configuration intents from the Conversational Agent and orchestrates on-chain execution of vault operations according to protocol rules. Acting as the downstream executor in the agent stack, it coordinates with vault smart contracts, risk policy layers, and asset deployment rails.

Responsibilities

1. Action Packet Reception

   1. Accepts structured instruction objects from the Conversational Agent or planning layer, such as "deposit 10K into vault ptmax," "stake in atvUSDC," or "withdraw amount from 808."

   2. Validates each packet against current policy boundaries, including risk bands, allocation ceilings, and asset eligibility.

2. Pre-Execution Checks

   1. Queries smart contract state and external oracles (using addresses listed in documentation) for current TVL, liquidity, yield differential, gas cost, and chain status.

   2. Confirms that the intended action will not violate diversification, slippage, or liquidity constraints defined in the protocol's Risk & Policy layer.

   3. If a violation is detected, the Execution Agent either rejects the action or triggers a clarification loop via the Conversational Agent.

3. Execution Orchestration

   1. Once approved, the Execution Agent prepares the transaction bundle:

      1. Encodes the required smart contract calls (vault deposit/withdraw/strategy update).

      2. Selects the correct chain (Ethereum, Arbitrum, Base).

      3. Routes the transaction through a gas-optimized path or MEV-resistant route if configured.

   2. Submits the transaction to the blockchain, monitors for inclusion, and handles failure remediation or retry attempts.

   3. Logs transaction metadata for audit trail and governance reporting.

4. Post-Execution Monitoring & State Update

   1. Reads on-chain receipts and updates internal system state, including vault share counts, user wallet exposure, and strategy allocations.

   2. Notifies the Conversational Agent and user of the outcome: success, partial fill, or failure/rollback.

   3. Flags execution anomalies: such as slippage above threshold or higher-than-expected gas costs: for risk alerts or governance review.

Integration & Interfaces

• Input Interface: Receives action packets from the Planning/Validation layer, driven by user intent via the Conversational Agent. Packets include structured JSON-like objects specifying vault, asset, amount, chain, and metadata.

• Policy Lookup Interface: Calls policy/risk smart contracts or on-chain modules to validate action feasibility (see policy & risk layer in documentation).

• Smart Contract Interface: Interacts with ERC-4626 compliant vault contracts and strategy modules using the contract addresses published in documentation.

• Notification Interface: Emits events or messages back to the system for the Conversational Agent and UI layer, such as "deposit complete," "rebalance executed," or "action failed."

Key Properties & Constraints

• Determinism: Execution must respect protocol-defined rules, including allocation bands, diversification limits, and timelocks. No subjective overrides are permitted outside governance.

• Non-custodial & Transparent: All transactions are on-chain, verifiable, and executed via audited smart contracts. Users retain full control of their funds.

• Auditable Trail: Every execution call is logged, with reasoning linked (via agent metadata) to the original intent, enabling full traceability.

• Cross-chain Support: Supports deployment across multiple EVM chains (Ethereum, Base, Arbitrum). Correct chain routing is essential for cost efficiency and access.

• Failure Handling: The Execution Agent implements fallback and retry logic for blockchain issues, including gas spikes, failed calls, and network congestion.

Latency & Slippage Control: For market-sensitive vaults (e.g., dynamic rebalancing), execution minimizes delay, slippage, and MEV exposure.