ERC-8004 feedback forensics skill for OpenClaw agents.
Most reputation systems show you the rating. This one tells you if the rating is real.
Before your agent pays another agent, this skill checks:
- Sybil risk: Are the reviews from real agents or fake wallet clusters?
- Reviewer quality: Is a 5-star rating from trusted agents or throwaway accounts?
- Cross-chain laundering: Does the agent have different reputations on different chains?
- Temporal anomalies: Sudden review bursts, rating manipulation patterns
- Spam filtering: Known spam feedbacks detected via tag analysis and excluded from scoring
Covers agents across 20 chains including Base, Ethereum, BSC, Celo, Gnosis, Optimism, Arbitrum, Avalanche, Polygon, Monad, and Solana. Call /stats for live counts.
Paste this repo URL into your OpenClaw chat:
https://github.com/goatgaucho/trustlayer-sybil-scanner
Or manually copy SKILL.md to your OpenClaw skills directory.
Ask your agent:
- "Check if agent base:1378 is legit"
- "Scan base:5000 for Sybil risk before I pay them"
- "Are the reviews for ethereum:42 real?"
- "Show me the most trusted agents on Base"
No API key needed (beta). x402 micropayments on paid endpoints ($0.001 USDC per query). All endpoints at https://api.thetrustlayer.xyz
| Endpoint | What it does |
|---|---|
/trust/<chain>:<agentId> |
Full Sybil scan + trust score |
/agent/<chain>:<agentId> |
Full agent profile, metadata, on-chain registration (paid) |
/reviewer/<wallet> |
Reviewer quality score + history (paid) |
/owner/<wallet> |
All agents owned by a wallet across chains (paid) |
/history/<chain>:<agentId> |
Daily score time-series + volatility (paid) |
/leaderboard?chain=base |
Top agents, Sybil-filtered |
/stats |
Live network-wide statistics |
Visual reports: https://thetrustlayer.xyz/agent/<chain>:<agentId>
ERC-8004 gives agents on-chain ratings. But a 4.5-star agent with reviews from a Sybil ring is not a 4.5-star agent. The spec authors anticipated "specialized reputation aggregators" would provide advanced scoring intelligence. That's what this is.
Six detection methods. Daily updates. 90-day score history retained.
Built by @GoatGaucho