Fraud & Safety Agents¶

Seven agents that handle everything from a quick fraud check on a single wallet to batch Sybil detection and autonomous transaction screening in a live AI pipeline.

chainaware-wallet-auditor¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_behaviour, WebFetch, WebSearch

The most comprehensive single-wallet analysis agent. Calls predictive_behaviour for a full intelligence report - fraud probability, experience, risk profile, and DeFi activity patterns - and uses WebFetch/WebSearch for external context when needed.

Use this when you need the full picture - not just a pass/fail, but the reasoning behind it.

Example prompts¶

Full audit of 0x742d35Cc6634C0532925a3b8D4C9C4e5B7e1234 on Ethereum

Due diligence on this wallet before we approve their grant application:
0xAbcD1234... on BNB Smart Chain

Complete analysis of the deployer for contract 0x9f8F72... on Base

Deep dive on this address - they want to LP $200k into our pool: 0x1a2b3c...

Example output¶

WALLET AUDIT REPORT
Address: 0x742d35Cc... | Network: Ethereum

FRAUD ASSESSMENT
  Status: Not Fraud
  Probability: 0.04 (Very Low)
  Forensic flags: None

BEHAVIOUR PROFILE
  Experience: Veteran (score: 0.91)
  Risk Willingness: Moderate (0.58)
  Risk Capability: High (0.82)
  Primary categories: DeFi Lending (41%), DEX Trading (33%), Bridging (18%)
  Protocols: Aave, Uniswap, Compound, Curve, Stargate

RUG PULL RISK (as deployer)
  Probability: 0.07 - Low risk

REPUTATION SCORE: 3,847 / 4,000

RECOMMENDATION: ✅ Proceed. Veteran DeFi participant with clean history.

chainaware-fraud-detector¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_fraud, predictive_fraud_batch, check_job_status, get_job_results

Fast, lightweight fraud screening. Returns fraud status, probability, and any forensic AML flags. Runs in under a second.

Use this for high-volume screening where you need a quick yes/no before proceeding to deeper analysis.

Example prompts¶

Is this wallet safe? 0x3fC91A3afd70395Cd496C647d5a6CC9D4B2b7FAD on Ethereum

Fraud check on 0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045 before I send ETH

AML screen: 0xBE0eB53F46cd790Cd13851d5EFf43D12404d33E8 on BNB Chain

Is this address suspicious? Someone sent me a payment request from 0x1234...

Quick check before I transact - 0xabc... on Polygon

Example output¶

FRAUD CHECK: 0x3fC91A3afd70395...
Network: Ethereum

Status: ⚠️ HIGH FRAUD RISK
Probability: 0.87

Forensic flags:
  - Connected to known phishing cluster (3 hops)
  - Mixer interaction (Tornado Cash, 45 days ago)
  - Exit scam counterparty (2 known rug pulls)

RECOMMENDATION: 🛑 Do not transact.

FRAUD CHECK: 0xd8dA6BF26964...
Network: Ethereum

Status: ✅ Not Fraud
Probability: 0.02

Forensic flags: None

RECOMMENDATION: ✅ Safe to proceed.

chainaware-rug-pull-detector¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_rug_pull, predictive_fraud, WebSearch

Smart contract and liquidity pool safety screening. Analyses the deployer wallet's behavioural history and LP structure to predict whether a project will rug.

Effective on Ethereum, BNB Smart Chain, Base, and Haqq. Most useful in the first hour after a pool launches, when social signals don't yet exist.

Example prompts¶

Will this pool rug pull? 0xPairAddress... on BNB Smart Chain

Is this contract safe to ape into? 0xTokenContract... on Ethereum

Check this LP before I add $10k liquidity: 0x... on Base

Rug pull risk on the deployer of 0xNewToken... on BSC

Should I ape in? Just launched an hour ago: 0x... on ETH

Example output¶

RUG PULL ASSESSMENT
Contract: 0xPairAddress... | Network: BSC

Risk Score: 0.83 - 🔴 HIGH RISK

Deployer wallet: 0xAbcd1234...
  Fraud probability: 0.79
  History: 2 previous rug pulls confirmed
  Wallet age: 11 days (no prior history)

LP Analysis:
  LP concentration: Single address holds 94% of LP
  LP wallet age: 3 days
  LP fraud score: 0.71

RECOMMENDATION: 🛑 Avoid. Deployer has a confirmed rug history.
                LP is concentrated and held by a fresh high-risk wallet.

RUG PULL ASSESSMENT
Contract: 0xEstablished... | Network: Ethereum

Risk Score: 0.12 - 🟢 LOW RISK

Deployer wallet: 0xVeteran...
  Fraud probability: 0.03
  History: 3 years on-chain, multiple successful projects

RECOMMENDATION: ✅ Low risk. Experienced deployer with clean record.

chainaware-counterparty-screener¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_fraud, predictive_behaviour

Real-time pre-transaction safety check. Returns a three-tier verdict - Safe / Caution / Block - before you send funds, confirm a trade, or interact with a contract.

Designed for speed: one address in, one verdict out.

Example prompts¶

Is it safe to send 5 ETH to 0x742d35... on Ethereum?

Check this counterparty before I confirm the trade: 0xAbcd... on BNB Chain

Pre-transaction check on 0x1234... - they want to swap tokens with me

Quick safety check on 0x... before I deposit into their LP

Should I trade with this wallet? They DM'd me about a deal: 0x...

Example output¶

COUNTERPARTY CHECK
Address: 0x742d35... | Network: Ethereum | Type: Transfer

Verdict: 🟢 SAFE

Fraud probability: 0.03
Experience: High (Veteran)
AML flags: None

Recommended action: Proceed normally.

COUNTERPARTY CHECK
Address: 0xAbcd... | Network: BNB Chain | Type: Trade

Verdict: 🔴 BLOCK

Fraud probability: 0.91
Key signals:
  - Known wash trading history
  - Connected to 4 confirmed exit scams
  - Fresh wallet (7 days old) with large balance

Recommended action: Do not proceed. High probability of fraud.

chainaware-trust-scorer¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_fraud

Returns a single trust score between 0.00 and 1.00 (1 - fraud_probability). The simplest possible interface for yes/no trust decisions.

Use this when you want a number you can compare, threshold, or display in a UI.

Example prompts¶

Trust score for 0x742d35Cc... on Ethereum

How much do I trust this wallet? 0xAbcd1234... on BNB Chain

Trustworthiness rating for 0x... - I need a number

Give me confidence scores for these 5 wallets: [list]

Example output¶

Trust Score: 0.97 ✅
Address: 0x742d35Cc...
Network: Ethereum
(Fraud probability: 0.03)

Trust Score: 0.09 🛑
Address: 0xBadActor...
Network: Ethereum
(Fraud probability: 0.91)

chainaware-transaction-monitor¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_fraud, predictive_rug_pull, predictive_behaviour

Built for autonomous AI agents and automated pipelines. Screens sender, receiver, and contract address simultaneously, then returns a composite risk score (0-100) and a machine-actionable pipeline action.

Four actions: ALLOW / FLAG / HOLD / BLOCK.

Example prompts¶

Should my agent execute this transaction?
Sender: 0xSender... Receiver: 0xReceiver... on Ethereum, value: 2.5 ETH

Risk score for this pending swap:
From 0x... to contract 0x... on Base - 1000 USDC

Monitor this transaction before I broadcast it:
sender=0x... receiver=0x... contract=0xUniswap... type=swap

Flag or allow this transfer?
0xUser... → 0xBridge... on BNB Chain, bridging $50k

Autonomous screening - batch check these 10 pending transactions: [list]

Example output¶

TRANSACTION RISK REPORT
Sender:    0xSender...   Risk: LOW (0.04)
Receiver:  0xReceiver... Risk: LOW (0.02)
Contract:  0xUniswap...  Risk: LOW (0.01)

Composite Risk Score: 8 / 100

Pipeline Action: ✅ ALLOW
Operator note: All parties clean. Standard swap. No intervention needed.

TRANSACTION RISK REPORT
Sender:    0xUser...     Risk: LOW  (0.05)
Receiver:  0xFlagged...  Risk: HIGH (0.88)

Composite Risk Score: 79 / 100

Pipeline Action: 🛑 BLOCK
Operator note: Receiver has confirmed fraud history.
               Do not relay this transaction.

chainaware-sybil-detector¶

Model: claude-haiku-4-5-20251001 | Tools: predictive_behaviour, predictive_fraud, predictive_behaviour_batch, predictive_fraud_batch, check_job_status, get_job_results

Batch Sybil detection via pattern recognition. Identifies wallet farms, coordinated airdrop attacks, and proxy voting fraud by analysing behavioural fingerprints across a list of addresses. Returns a per-wallet verdict and an overall Sybil cluster map.

Use this before airdrops, IDOs, governance votes, or any event where one actor controlling many wallets creates a fairness or security problem.

Example prompts¶

Check these 500 airdrop applicants for Sybil wallets: [list] on Ethereum

Are any of these DAO voters the same person? [addresses] on BNB Chain

Find wallet farms in this list - we're about to distribute tokens: [list]

Sybil check on our IDO whitelist before we finalise allocations: [list]

Example output¶

SYBIL DETECTION REPORT
Input: 500 wallets | Network: Ethereum | Goal: Airdrop screening

SUMMARY
  Clean wallets:         431 (86%)
  Suspected Sybils:       54 (11%)
  Confirmed bot farms:    15 (3%)

CLUSTER MAP
  Cluster A: 12 wallets - same funding source, created within 4 hours
  Cluster B: 8 wallets  - identical transaction patterns, shared IP signal
  Cluster C: 19 wallets - all funded from the same CEX withdrawal batch

RECOMMENDATION
  Remove 69 addresses (clusters A, B, C + 15 confirmed bots).
  Reallocate their share proportionally to clean wallets.

Fraud & Safety Agents¶

chainaware-wallet-auditor¶

Example prompts¶

Example output¶

chainaware-fraud-detector¶

Example prompts¶

Example output¶

chainaware-rug-pull-detector¶

Example prompts¶

Example output¶

chainaware-counterparty-screener¶

Example prompts¶

Example output¶

chainaware-trust-scorer¶

Example prompts¶

Example output¶

chainaware-transaction-monitor¶

Example prompts¶

Example output¶

chainaware-sybil-detector¶

Example prompts¶

Example output¶

Further Reading¶