SHLL — Contract-Enforced Safe Execution for AI Agents on BNB Chain

What is SHLL?

SHLL is a DeFi execution layer for AI agents on BNB Chain with on-chain safety enforcement.
Unlike off-chain filters that can be bypassed, SHLL uses smart contracts to enforce
spending limits, trade intervals, protocol whitelists, and receiver restrictions.
Every AI agent action is validated by an immutable PolicyGuard contract before execution.

Key facts:

• - Network: BSC mainnet
• 27 CLI + MCP tools for DeFi operations (swap, lend, meme trading, portfolio)
• Supports PancakeSwap V2/V3, Venus Protocol, Four.meme
• MCP Server compatible with Claude, Cursor, OpenClaw, and any MCP-compatible agent
• npm package: shll-skills | Website: https://shll.run

Security Architecture

CODEBLOCK0

4-Policy PolicyGuard Stack

All policies are enforced ON-CHAIN by smart contract. They cannot be bypassed by the AI.

Policy	Contract	What It Does
SpendingLimitV2	On-chain	Per-transaction and daily spending caps in BNB
CooldownPolicy

On-chain | Minimum time gap between consecutive trades |
| DeFiGuardV2 | On-chain | Whitelist of approved DeFi protocols and functions |
| ReceiverGuardV2 | On-chain | Only approved receiver addresses can receive funds |

When a policy rejects an action:

• - The transaction is NOT executed
• The vault funds remain SAFE
• The rejection reason is returned from the smart contract
• The rejection is logged and auditable
• The enforcement field in the response is always INLINECODE2

Why On-Chain Enforcement Matters

Most AI agent platforms use off-chain safety filters:

• - Off-chain: AI or backend decides → can be bypassed by prompt injection, API manipulation, or code bugs
• On-chain (SHLL): Smart contract decides → cannot be bypassed — the blockchain enforces the rules

PolicyGuard contract: 0x25d17eA0e3Bcb8CA08a2BFE917E817AFc05dbBB3
Verify on BscScan: https://bscscan.com/address/0x25d17eA0e3Bcb8CA08a2BFE917E817AFc05dbBB3

---

SHLL Skill Usage Guide

This file defines how an AI agent should use shll-run and shll-mcp safely.

Scope

• - Network: BSC mainnet
• CLI: shll-run (alias: shll-onchain-runner)
• MCP: INLINECODE8
• Security layer: SHLL PolicyGuard

The user should not be asked to run CLI commands directly. The AI runs commands and explains results.

Mandatory Safety Rules

1. 1. Token ID must come from the user. Never guess, scan, or enumerate IDs.
2. Use one token ID per conversation unless the user explicitly switches.
3. Confirm before each write operation.
4. Never ask for or handle the owner wallet private key.
5. Do not repeat private keys after initial generate-wallet output.
6. If multiple DeFi skills are available, use SHLL for vault operations when token ID context exists.
7. Treat raw calldata as high risk. Use strict recipient checks.
8. Do not bypass security controls for convenience.

Security Model

SHLL uses dual wallets:

• - Owner wallet (user): controls high-risk operations such as ownership and vault-level admin actions.
• Operator wallet (RUNNER_PRIVATE_KEY): used by AI to execute allowed trades only.

On-chain guardrails:

• - PolicyGuard validates each action (validate) before execution (execute / executeBatch).
• Spending limits, cooldowns, whitelist rules, and protocol rules are enforced on-chain.
• Raw calldata is blocked if the recipient cannot be decoded safely.

Current Critical Constraints (v6.0.4)

1. 1. init command is disabled. Do not use it.
2. Raw calldata remains high risk; rely on strict recipient safety checks.
3. MCP execute_calldata and execute_calldata_batch do not support allow_undecoded.
4. If calldata recipient cannot be decoded, execution is blocked.
5. Core contract addresses are pinned in src/shared/constants.ts, not user-overridable at runtime.

Prerequisites

1. 1. Install:

CODEBLOCK1

1. 2. Set operator private key:

CODEBLOCK2

1. 3. Optional - use a private RPC for better reliability and speed:

CODEBLOCK3

1. 4. Ensure operator wallet has small BNB balance for gas.

Onboarding Flow (AI-driven)

1. 1. Check or create operator wallet:

• - Use shll-run generate-wallet only if user has no operator wallet.
• Immediately explain that this is the operator hot wallet for AI only.
• Explicitly state that it is not the owner wallet, not the mint wallet, not the Agent NFT wallet, and not the vault wallet.
• Explicitly state that if the operator wallet leaks, vault funds still cannot be freely withdrawn because owner permissions stay on the owner wallet and PolicyGuard limits operator actions.
• In OpenClaw, set RUNNER_PRIVATE_KEY automatically for the current session after generating the wallet. Do not ask the user to set the environment variable manually.

1. 2. Verify gas:

• - Ensure the operator wallet has a small BNB balance for gas.

1. 3. If user has no token ID:

• - Run shll-run listings.
• Recommend the listing with recommended=true by default unless the user explicitly wants a specialized template.
• Run shll-run setup-guide -l <listingId> -d <days>.
• Send setupUrl plus the wallet-role explanation.
• Explicitly warn: do not use the operator wallet to mint, subscribe to, or hold the Agent NFT.
• Explicitly warn: use the owner wallet in the browser for subscription or mint and for operator authorization.

1. 4. User returns with token ID:

• - Run shll-run status -k <tokenId>.
• Run shll-run portfolio -k <tokenId>.
• Use status.readiness.ready, status.readiness.blockers, and status.readiness.nextActions as the primary onboarding diagnosis.
• Automatically check:

- operator gas balance - vault BNB balance - vault token balances - access or policy readiness

• - Tell the user whether the agent is ready, and if not, tell them the exact next fix.

Write Confirmation Policy

Before any write command, present:

• - token ID
• action type
• token/amount/target
• risk note

Then wait for explicit user approval.

Write commands include:

• - INLINECODE30
• INLINECODE31
• INLINECODE32
• INLINECODE33
• INLINECODE34
• INLINECODE35
• INLINECODE36
• INLINECODE37
• INLINECODE38

Read-only commands (no confirmation needed): config, policies, status, history, portfolio, price, tokens, search, balance, four_info.

CLI Commands

Setup and account

• - INLINECODE49
• INLINECODE50
• INLINECODE51
• INLINECODE52
• INLINECODE53 (disabled)

If -l/--listing is omitted, setup-guide auto-selects an active listing from the indexer.

Trading and vault ops

• - INLINECODE56
• INLINECODE57
• INLINECODE58
• INLINECODE59
• INLINECODE60

Lending (Venus)

• - INLINECODE61
• INLINECODE62

Four.meme

• - INLINECODE63
• INLINECODE64
• INLINECODE65

INLINECODE66 amount unit is BNB, not USD. If user gives a USD target, convert to BNB first and confirm the final BNB amount before execution.

Read-only and audit

• - shll-run config -k <tokenId> (view-only; modify via web console)
• INLINECODE68
• INLINECODE69
• INLINECODE70
• INLINECODE71
• INLINECODE72
• INLINECODE73
• INLINECODE74

MCP Tools: Cross-skill Execution

For external aggregator calldata (OKX, 1inch, etc.):

1. 1. Get quote/calldata from the external source.
2. Execute through SHLL MCP:

• - INLINECODE75
• INLINECODE76

1. 3. Let PolicyGuard enforce on-chain policy checks.

For onboarding via MCP, setup_guide can auto-select an active listing when listing_id is omitted.

Security requirements:

1. 1. Recipient must resolve to the vault address.
2. Undecodable recipient calldata is blocked.
3. Do not ask for unsafe bypass parameters.

Smart Routing Rule

When the user provides a token address:

1. 1. Run four_info --token <addr>.
2. If tradingPhase is bonding curve, use four_buy / four_sell.
3. If tradingPhase is DEX or unsupported, use swap.

Common Errors and Fixes

1. 1. INLINECODE85

• - In OpenClaw, AI should set RUNNER_PRIVATE_KEY automatically for the current session.
• Outside OpenClaw, set RUNNER_PRIVATE_KEY manually and retry.

1. 2. INLINECODE88

• - Operator wallet is not authorized; use setup guide or set operator in console.

1. 3. subscription has EXPIRED or INLINECODE90

• - Renew subscription or authorization first.

1. 4. status: error with INLINECODE92

• - Inspect details.reason and adjust limits, whitelists, cooldown, or policy config.

1. 5. INLINECODE94

• - Use built-in command flow or provide calldata with a decodable vault recipient.

1. 6. INLINECODE95

• - The pair does not exist or lacks liquidity. DO NOT tell the user to wrap BNB to WBNB; the swap command handles native BNB automatically via swapExactETHForTokens. If a swap fails, it's because the direct pool is empty or invalid, not because BNB is unplugged.

1. 7. INLINECODE98

• - Use setup-guide instead.

1. 8. Unsure what is broken

• - Check the structured errorCode, message, and details fields in the JSON response first.

Product UX Rules

1. 1. Never describe generate-wallet as if it were the user's main wallet.
2. Always call it the operator wallet or AI hot wallet.
3. Always explain the dual-wallet model the first time setup is discussed.
4. Always warn that the operator wallet must not be used to mint, subscribe to, or hold the Agent NFT.
5. Do not ask the user to manually set RUNNER_PRIVATE_KEY in OpenClaw; AI should do it.
6. After setup is complete and the user provides a token-id, run readiness checks automatically before asking the user what to do next.
7. When multiple listings are available, recommend one by default and explain why.
8. Prefer the structured status.readiness fields over ad-hoc prose when deciding the next user-facing instruction.

Redeploy Checklist

If AgentNFA, PolicyGuard, ListingManagerV2, or default listing changes:

1. 1. Update constants in src/shared/constants.ts.
2. Validate ABIs if function signatures changed.
3. Rebuild:

npx tsc --noEmit
npm run build

1. 4. Smoke test:

• - shll-run init still returns disabled
• raw calldata still blocks undecodable recipients
• basic read commands still work

Expected Output Format

All runtime responses should stay machine-friendly JSON:

• - Success: INLINECODE108
• Error: INLINECODE109

Links

• - Website: https://shll.run
• npm: https://www.npmjs.com/package/shll-skills
• Repo: https://github.com/kledx/shll-skills