Risk & Safety Agents¶

Key Takeaways
- Seven agents covering counterparty screening, lending risk assessment, credit scoring (1-9 on Ethereum), token launch auditing, AI agent wallet screening, portfolio risk, and RWA investor compliance
- chainaware-lending-risk-assessor evaluates repayment likelihood from on-chain history - enabling undercollateralised DeFi lending decisions
- chainaware-credit-scorer returns a 1-9 credit score combining fraud probability, behavioural history, and cash flow analysis
- chainaware-token-launch-auditor runs three parallel checks: rug pull risk on contract, fraud on deployer, and deployer behavioural history

Seven agents for evaluating risk across counterparties, lending positions, credit decisions, token launches, AI agent interactions, portfolios, and real-world asset investments.

Setup required: For AI Agents - MCP registration and agent installation.

chainaware-counterparty-screener¶

Role: Screens counterparties before executing trades, OTC deals, or B2B crypto transactions.

What it does: Evaluates both fraud risk and on-chain behaviour for a counterparty wallet before a business-to-business transaction is committed. Flags wallets with fraud history, unusual behavioural patterns (wash trading, sandwiching, Sybil indicators), or recent interaction with flagged contracts.

Tools used: predictive_fraud, predictive_behaviour
Model: Claude Haiku 4.5

Example invocation:

@chainaware-counterparty-screener Screen counterparty wallet 0xOTC...DEAL before we execute

Output includes:
- Fraud risk score and classification
- Behavioural profile summary
- Specific counterparty risk flags
- Proceed / Proceed with caution / Decline recommendation

chainaware-lending-risk-assessor¶

Role: Assesses borrower risk for undercollateralised DeFi lending decisions.

What it does: Analyses a borrower wallet's behavioural history (predictive_behaviour) and fraud signals (predictive_fraud) to produce a credit-risk-style assessment. Evaluates repayment likelihood based on past on-chain activity, DeFi engagement depth, and wallet tenure. Designed for protocols offering undercollateralised or reputation-based lending.

Tools used: predictive_behaviour, predictive_fraud, credit_score
Model: Claude Haiku 4.5

Example invocation:

@chainaware-lending-risk-assessor Assess lending risk for borrower wallet 0xBORROWER...999

Output includes:
- Borrower risk tier (Prime / Standard / Subprime / Decline)
- Behavioural indicators supporting the assessment
- Fraud risk overlay
- Recommended LTV ratio or loan limit
- Conditions or monitoring requirements

chainaware-token-launch-auditor¶

Role: Pre-launch and post-launch token risk audit combining deployer, token, and holder signals.

What it does: Runs a three-signal audit: predictive_rug_pull for the token contract itself, predictive_fraud for the deployer wallet, and predictive_behaviour for the deployer's historical on-chain activity. Produces a structured audit report suitable for community due diligence, launchpad listing decisions, or investor research.

Tools used: predictive_rug_pull, predictive_fraud, predictive_behaviour
Model: Claude Haiku 4.5

Example invocation:

@chainaware-token-launch-auditor Audit this new token launch: contract 0xTOKEN, deployer 0xDEPLOYER

Output includes:
- Token rug pull risk score
- Deployer fraud score
- Deployer behavioural profile (serial launcher? prior exits?)
- Combined launch risk verdict (Safe / Caution / High Risk / Avoid)
- Specific red flags by category
- Recommended next steps for investors

chainaware-agent-screener¶

Role: Screens AI agent wallets before allowing them to interact with a DeFi protocol or treasury.

What it does: As AI agents increasingly transact on-chain autonomously, protocols need to screen agent wallets for fraud patterns and unusual behaviours. This agent evaluates an on-chain AI agent wallet using predictive_fraud and predictive_behaviour, specifically looking for patterns associated with drain attacks, front-running bots, or compromised agent wallets.

Tools used: predictive_fraud, predictive_behaviour, predictive_rug_pull
Model: Claude Haiku 4.5

Example invocation:

@chainaware-agent-screener Screen this AI agent wallet before granting treasury access: 0xAGENT...WALLET

Output includes:
- Fraud risk assessment for the agent wallet
- Behavioural anomaly flags (bot-like patterns, MEV activity)
- Agent risk classification
- Allow / Restricted access / Deny recommendation
- Suggested permission scope if allowed

chainaware-portfolio-risk-advisor¶

Role: Portfolio-level risk advisory combining individual token quality with rug pull signals.

What it does: Takes a list of token holdings and evaluates overall portfolio risk by running token_rank_single (quality and rank signals) and predictive_rug_pull (rug pull probability) for each position. Aggregates into a portfolio risk score and identifies which holdings represent the greatest concentration of risk.

Tools used: predictive_rug_pull, token_rank_single
Model: Claude Haiku 4.5

Example invocation:

@chainaware-portfolio-risk-advisor Assess my portfolio risk: [0xTOKEN1, 0xTOKEN2, 0xTOKEN3]

Output includes:
- Per-token risk score and rank
- Portfolio overall risk rating
- Highest-risk positions ranked
- Concentration risk warnings
- Diversification recommendations
- Specific tokens to monitor or exit

chainaware-rwa-investor-screener¶

Role: Screens investors in Real-World Asset (RWA) tokenisation platforms for compliance and behaviour.

What it does: RWA platforms serving regulated investment products need to know that investors are not only KYC-compliant but also behave consistently with legitimate investment intent. This agent combines predictive_fraud (compliance risk) with predictive_behaviour (investment behavioural patterns) to screen wallets for RWA platform onboarding.

Tools used: predictive_fraud, predictive_behaviour
Model: Claude Haiku 4.5

Example invocation:

@chainaware-rwa-investor-screener Screen this investor wallet for RWA platform onboarding: 0xINVESTOR...ABC

Output includes:
- Fraud / AML risk classification
- Investment behavioural profile (institutional patterns, retail speculation, arbitrage)
- Suitability assessment for regulated RWA products
- Onboarding recommendation (Approve / Enhanced due diligence / Decline)
- Compliance notes for audit trail

chainaware-credit-scorer¶

Role: AI-driven crypto credit score for DeFi lending and credit decisions.

What it does: Calls credit_score to produce a numeric credit score (1-9) for a wallet address, combining fraud probability and social graph analysis. A score of 1 represents minimum creditworthiness; a score of 9 represents maximum. Designed for undercollateralised lending protocols, buy-now-pay-later crypto products, and any DeFi mechanism that extends credit based on on-chain reputation rather than collateral alone.

Tools used: credit_score
Model: Claude Haiku 4.5
Networks: Ethereum

Example invocation:

@chainaware-credit-scorer What is the credit score for wallet 0xBORROWER...123?

Output includes:
- Credit score (1-9)
- Credit tier (Prime / Near-prime / Subprime / Decline)
- Key factors driving the score (behavioural signals, fraud probability)
- Recommended credit limit or LTV tier
- Monitoring notes

Frequently Asked Questions¶

How does the lending risk assessor differ from the credit scorer?
chainaware-lending-risk-assessor produces a narrative risk assessment with a borrower tier and recommended LTV, drawing on predictive_behaviour, predictive_fraud, and credit_score. chainaware-credit-scorer returns only the numeric 1-9 score with a tier label - it is the simpler, faster call for protocols that just need a number to gate on.

Does the token launch auditor work on already-launched tokens or only pre-launch?
It works on both. For pre-launch tokens you pass the contract and deployer addresses and receive a pre-investment audit. For already-launched tokens you pass the live contract address and the agent retrieves the current risk picture from on-chain data.

Can the counterparty screener flag an entire counterparty organisation, not just a single wallet?
The screener evaluates one wallet address per call. For organisations with multiple wallets, invoke it once per address and combine the results. chainaware-compliance-screener can orchestrate batch screening of multiple counterparty wallets in one step.

Further Reading¶

• Web3 Trust Verification Systems in 2026
• ChainAware's 32 Claude Sub-Agents - Fraud Tech and Growth Tech