convez.ai
AgenteSegurançaAvançado
PurpleAILAB

Vigilo

PurpleAILAB/Vigilo

Legião de IA para auditoria autônoma de smart contracts Web3

Ver no GitHub
BomPluginScore · de 0 a 100

Agent autônomo para auditoria de segurança em smart contracts Web3, utilizando uma legião de especialistas (inspetores, exploradores e auditores) para identificar vulnerabilidades e gerar provas de conceito validadas. Projetado para bug bounties, competições de auditoria e pesquisa ofensiva.

Instalação

$npx convez-ai PurpleAILAB/Vigilo

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
73
Forks
23
Acompanhando
1
Saúde
62%
Último commit
há 3 semanas
Versão
v1.0.2
Issues abertas
0
Contribuidores
3

Casos de uso

  • Auditoria automática de smart contracts em competições Cantina e similares
  • Identificação de vulnerabilidades críticas em protocolos DeFi
  • Pesquisa ofensiva e desenvolvimento de exploits em Web3

Veredito

Ferramenta especializada e comprovada em produção que automatiza auditoria de smart contracts através de multi-agentes com papéis específicos. Adequada para auditores e desenvolvedores de protocolos Web3 que precisam de análise sistemática de vulnerabilidades.

README

A tradução para o português deste README ainda está na fila.

npm version GitHub Release GitHub Stars GitHub Issues License


What is Vigilo?

Vigilo is an autonomous security legion for smart contract auditing, inspired by the command structure of the Roman army. It runs inside OpenCode, deploying specialized agents in parallel to find vulnerabilities and generate validated PoCs.

The Legion

AgentMission
VigiloOrchestrates the full audit pipeline
QuaestorPre-audit interview & scope planning
ExploratorCode reconnaissance — maps structure and flows
SpeculatorDocumentation intel — extracts design and invariants
Centuriones8 specialist auditors deployed by protocol type

Proven in the Wild

Real vulnerabilities found by Vigilo in live audit contests.

All I did was specify the target project and submit the report. Vigilo did the rest.


Installation

OpenCode

For LLM Agents (Recommended)

Paste this into your LLM agent session:

Install and configure vigilo by following the instructions here:
https://raw.githubusercontent.com/PurpleAILAB/Vigilo/main/packages/opencode/docs/installation.md

Manual Install

bunx vigilo install

Claude Code

/plugin marketplace add PurpleAILAB/Vigilo
/plugin install vigilo@Vigilo

See the full Installation Guide for more options.

Uninstallation

  1. Remove the plugin from your OpenCode config:
# Edit ~/.config/opencode/opencode.json and remove "vigilo" from the plugin array
  1. Remove configuration files:
rm -f ~/.config/opencode/vigilo.json
  1. Verify removal:
opencode --version

Features

  • Automated Audit Workflow: Scope → Recon (Exploratores) → Deep Analysis (Centuriones) → PoC → Report
  • Specialized Auditors: Reentrancy, Oracle, Access Control, Flashloan, Logic, DeFi, Token, Cross-Chain
  • Multi-Language Support: Solidity, Vyper, Cairo, Rust
  • Foundry Integration: forge build, forge test, forge coverage
  • LSP Integration: Goto-definition, references, diagnostics
  • Parallel Analysis: Multiple auditors running concurrently
  • PoC Validation: Auto-generate and validate Foundry tests

Usage

cd my-solidity-project
opencode

# Start audit
/audit

# Generate PoC
/poc .vigilo/findings/high/H-01-reentrancy.md

Directory Structure

.vigilo/
├── recon/           # Explorator & Speculator outputs
├── findings/        # Vulnerability findings
│   ├── high/
│   └── medium/
├── poc/             # PoC validation logs
└── reports/         # Final reports

Platforms

PlatformPackageStatus
OpenCodepackages/opencode⭐ Recommended
Claude Codepackages/claudeStable

Why OpenCode? More flexibility with model selection, better plugin extensibility, and cost-effective auditing with configurable models per auditor.


Benchmarking

Measure Vigilo's audit accuracy against verified security reports from Code4rena, Sherlock, and Cantina.

# Run full benchmark pipeline
bunx vigilo-bench sherlock_cork-protocol_2025_01 -w -v

Pipeline: checkout → audit → score → report

See packages/bench for full documentation.


Troubleshooting

bunx vigilo doctor
bunx vigilo doctor --verbose
IssueSolution
OpenCode not foundInstall from https://github.com/anomalyco/opencode
Foundry not foundcurl -L https://foundry.paradigm.xyz | bash && foundryup
Vigilo not registeredRun bunx vigilo install again

Development

For contributors working on Vigilo itself.

Setup

git clone https://github.com/PurpleAILAB/Vigilo.git
cd vigilo/packages/opencode
bun install
bun link

Development Mode

  1. Configure local plugin path in ~/.config/opencode/opencode.json:
{
  "plugin": [
    "D:/path/to/vigilo/packages/opencode"
  ]
}
  1. Run watch mode:
bun run dev
  1. Restart OpenCode to load changes.

Quick Commands

TaskCommand
Buildbun run build
Watch modebun run dev
Test CLIbun src/cli/index.ts install
Run doctorbun src/cli/index.ts doctor --verbose

Restore Production Mode

bunx vigilo install

This resets the plugin path to vigilo@latest.


License

Business Source License 1.1

  • Non-production use: Free
  • Production use: Requires commercial license
  • Change Date: 2029-01-21 (converts to Apache-2.0)

Commercial licensing: catower917@gmail.com


Ready to hunt bugs? 🔍

Get Started · Report Bug · Request Feature

Ver o original no GitHub

Bom para

Relacionados