Forge Protocol
Forge Protocol's Team
Problem Statement
Automated security tools can find bugs. But nobody can verify whether the tool that found (or missed) a vulnerability is actually reliable. Current AI-powered scanners (Copilot Autofix, Snyk Agent Fix) have no on-chain accountability, no verifiable track record, no mechanism for other agents to trust their results. When an automated auditor hallucinates a finding or misses a real vulnerability, there is zero accountability — because the tool has no persistent identity. Forge Protocol solves this: autonomous multi-agent security auditing where every agent has an ERC-8004 on-chain identity, every audit builds verifiable reputation, and agents verify each other trust before collaborating.
Forge Protocol is the first autonomous security auditor with on-chain accountability. No existing tool (Copilot Autofix, Snyk, Aardvark) can prove it audited reliably, build verifiable reputation, or let other agents trust its results on-chain. DEMO VIDEO: https://www.loom.com/share/ac1b0120ffc84d578a4524ed457c38a1?t=0 PROBLEM CLARITY: Automated security tools have zero accountability. They hallucinate findings, miss real vulnerabilities, and nobody knows until production breaks. There is no way to verify an auditors track record. TECHNICAL EXECUTION: Five specialized agents run a 6-step autonomous pipeline (Plan, Scan, Analyze, Fix, Review, Self-Correct). Five ground-truth tools: Semgrep 1.156.0, custom SAST (12 CWE rules), GitHub Advisory CVE Database, GitHub API, Claude AI. The system autonomously forks repos, commits SECURITY_AUDIT.md reports, and opens Pull Requests. Example: github.com/ElijahUmana/forge-protocol/pull/7 AI x CRYPTO INTEGRATION: ERC-8004 Agent #2221 on Ethereum Sepolia. Trust-gated collaboration via ownerOf(). Dynamic reputation from actual audit metrics. Independent Reviewer wallet. x402 micropayment headers. ORIGINALITY: No production system combines autonomous security auditing with on-chain identity and verifiable reputation. IMPACT: 14 real findings per audit. Autonomous PR creation delivers fixes directly. COMPLETENESS: Working end-to-end. Live dashboard, inter-agent communication, trust verification, PR links, AI summaries, session history.
Build Timeline
Team
Forge Protocol
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
Extending Forge Protocol with production SAST integration (Semgrep), multi-chain reputation aggregation, agent marketplace for hiring security auditors, and automated PR creation with verified fix deployment.