Autonomous Code Audit Agent
Agent Wallet Protocol's Team
Problem Statement
Security audits are expensive, slow, and bottlenecked by human availability. Most open-source repositories never get a professional security review. Automated scanners exist but require human setup, configuration, and interpretation. The challenge is building an agent that can autonomously discover repositories, plan appropriate audit strategies, execute comprehensive analysis, and produce actionable reports — all without human intervention. This agent demonstrates that AI can handle the full audit lifecycle autonomously while maintaining safety guardrails and compute budget awareness.
A fully autonomous AI agent that performs security audits on GitHub repositories without human intervention. Implements the complete autonomous loop: discover repos via GitHub API → plan audit scope based on codebase structure → execute multi-pattern static analysis → verify findings against known vulnerability databases → generate structured reports with severity ratings. Features ERC-8004 on-chain identity, a machine-readable agent.json capability manifest, and structured execution logs (agent_log.json). The agent enforces compute budget awareness (max iterations, max API calls) and includes comprehensive safety guardrails: input validation, unsafe operation detection, rate limiting, and abort/retry mechanisms. Scans for 6+ vulnerability classes: hardcoded secrets, SQL injection, XSS, eval/exec usage, prototype pollution, and command injection. 38 tests covering agent lifecycle, analysis accuracy, guardrails, logging, and reporting.
Build Timeline
Team
Agent Wallet 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
Intention
Plans to continue
Planning AST-based analysis, dependency vulnerability checking, multi-language support