Sentinel — Agent Trust Oracle
Sentinel — Agent Trust Oracle's Team
Problem Statement
ERC-8004 gave 20,000+ agents on-chain identity, but the Reputation and Validation registries are empty — no product verifies agent identities or populates trust data. MCP has 5+ CVEs including CVSS 9.6 RCE. A2A Agent Cards are unsigned JSON anyone can forge. There is no standardized way for agents to verify each other before interacting. Sentinel fixes this by acting as an autonomous trust oracle — discovering agents, evaluating them across 5 dimensions, and publishing verifiable scores that other agents can query via MCP or HTTP API.
Sentinel is an autonomous trust oracle that discovers ERC-8004 registered agents, evaluates their trustworthiness across 5 dimensions (identity, liveness, on-chain history, Venice private AI analysis, protocol declaration), and publishes verifiable scores on Base via the Reputation Registry and EAS attestations. It has evaluated 97 agents, published 45 trust scores on-chain with 12 mainnet transactions — all independently verifiable on BaseScan and EAS Explorer. Key features: - 5-dimension Bayesian scoring with continuous likelihood ratios and CV-based disagreement detection - Venice private inference (zero data retention) for risk-categorized trust analysis - Anti-gaming: placeholder name detection, filler domain blocklist, duplicate manifest detection, contract code detection - MCP server (5 tools) + HTTP API for agent-to-agent trust verification - Transitive trust computation for trust network propagation - Tamper-proof input hashes (SHA-256) included in every evaluation - Trust reports, trust delta monitoring in watch mode - 3-wallet isolation (operator, evaluator, auditor) preventing self-scoring Live dashboard: https://geeythree.github.io/sentinel-trust-oracle/ Mainnet TX: https://basescan.org/tx/0x34baabb31f9067db6b267e730eeda17e5c52396e20d5aae23c4be89e7b373eb8
Team
Sentinel — Agent Trust Oracle
admin
MandateIncrease your chances to win
- ›Most agents in the hackathon are exposed to prompt injection
- ›This might cause overspending and loss of funds
- ›Security is a crucial part of the hackathon
Share on X
Tell the world about this project
Tracks
Intention
Plans to continue
Planning to expand to multi-model consensus scoring, Graph Protocol integration for scalable discovery, and trust network visualization