[New Agent Template]: Radiant Strategy Pack (Looping Lender, HF Shield, Auto-Compounder)

[Lexiie]

🏷️ Agent Name

Radiant Strategy Pack

📝 Agent Description

The Radiant Strategy Pack is a community-contributed collection of 3 DeFi agent templates designed for the Vibekit ecosystem. These agents automate multiple Radiant lending strategies on Arbitrum and leverage the Radiant Lending Plugin requested in Issue #310.

This agent pack provides automated looping, health factor protection, and auto-compounding of RDNT rewards. It exposes a clean strategy interface and allows end-users to run each strategy through a CLI without writing code.

Included strategies:

1. Radiant Leveraged Looping Lender

   • Automates supply/borrow loops
   • HF-aware and borrow-cap aware
   • Safe stopping and guardrails

2. Radiant Health Factor Shield

   • Monitors HF
   • Performs soft deleverage
   • Triggers hard emergency exit when needed

3. Radiant Rewards Auto-Compounder

   • Claims RDNT
   • Swaps to target asset
   • Supplies back into Radiant

Key features:

• Built on top of the Radiant Lending Plugin
• Abstraction layer (RadiantClient) for clean interactions
• CLI runner for non-developers
• Full unit tests with mocks
• Safe execution guardrails
• Located in: typescript/community/agents/radiant-strategy-pack/

Implements and extends:
#310

🔌 Required Protocols

Primary protocols:

• Radiant V2 Lending Protocol on Arbitrum One
• Radiant Lending Onchain-Actions Plugin (markets, positions, actions)

Optional protocols:

• None. All data is fetched on-chain via viem and Radiant contracts without external API dependencies.

🤖 Automation Level

Manual - User triggers each action

🏗️ Code Architecture

Skills & Tools structure:

• Primary skills:
  • Fetch Radiant market data (APR, liquidity, prices)
  • Query user lending positions (collateral, debt, health factor)
  • Build lending transactions (supply, withdraw, borrow, repay, setCollateral)
• Internal tools:
  • RadiantClient abstraction used by the strategies
  • Radiant onchain-actions plugin: markets, positions, actions
  • CLI wrapper for running strategies without writing code
• MCP tools:
  • Not used in this agent. All tools are internal TypeScript modules.

Agent components:

• Context providers:
  • Wallet / signer configuration for sending transactions (outside this package – tx builders only return calldata)
  • RPC provider via viem for on-chain reads
• Hooks:
  • Logging hooks inside the strategies (loop iterations, HF checks, compound decisions)
  • Validation hooks for HF thresholds, borrow capacity, min compound threshold, and max loop limits
• Configuration & environment:
  • Per-strategy config (asset, HF thresholds, loop count, minRewards, etc.)
  • Radiant network config via radiant.config.ts (addresses, decimals, chainId)
  • viem client configuration for RPC URL

Integration patterns:

• All protocol interactions go through the Radiant onchain-actions plugin:
  • markets.ts for market data
  • positions.ts for user positions (HF, collateral, debt)
  • actions.ts for building supply/withdraw/borrow/repay/setCollateral txs
• Strategies never call contracts directly; they call RadiantClient, which wraps the plugin.
• Transaction builders return { to, data, value } only; the caller is responsible for signing & broadcasting transactions.
• Error handling and safety:
  • Looping strategy stops on low HF or when max loop count is reached.
  • Health Factor Shield strategy has soft, hard, and full-exit modes based on HF bands.
  • Auto-compounder skips if rewards are below a minimum threshold and uses a simplified 1:1 mock swap model for RDNT → target asset in the current version.

🧪 Testing Requirements

Unit tests:

• Radiant plugin:
  • markets.test.ts validates market data fetching.
  • positions.test.ts validates user position and health factor queries.
  • actions.test.ts validates tx builders for supply/withdraw/borrow/repay/setCollateral.
• Strategy pack:
  • radiantLooping.test.ts covers looping lender behaviour (HF stop, maxLoops, borrow/supply ordering).
  • healthShield.test.ts covers HF shield soft/hard deleverage and emergency exit.
  • compounder.test.ts covers reward threshold and compound path.
  • cli.test.ts covers CLI argument parsing and correct strategy dispatch, all using a mocked RadiantClient.

Integration tests:

• Local integration using mocked or forked RPC can be used to validate that:
  • Radiant plugin reads markets and positions correctly on Arbitrum.
  • Strategy pack composes plugin calls in the expected order.
• These are optional for merge but supported by the architecture.

Live testing:

• Designed to be run against Arbitrum mainnet or a mainnet fork:
  • User provides RPC URL and wallet with collateral.
  • CLI can be pointed to live Radiant markets.
• Live testing is not required for inclusion as a community agent template, due to collateral and risk constraints, but the flows are compatible with live deployment.

📎 Additional Information

Repository folder:
typescript/community/agents/radiant-strategy-pack/

This submission includes:

• Complete strategy pack
• Full strategy documentation
• CLI runner for easy user execution
• Unit test suite with mocks
• RadiantClient abstraction layer
• High-safety guardrails for DeFi automation

Implements the requirements from:
Issue #310 – Integrate Radiant Lending Protocol (Arbitrum)

✅ Pre-submission Checklist

• I have searched existing issues and templates to avoid duplicates
• I have clearly described the agent's purpose and functionality
• I have identified the required protocols and integrations
• I understand this agent will be created in the typescript/community/ directory
• I will wait for the Vibekit team to approve this issue before continuing to implementation