Skip to content

Quickstart

Run your first smart contract audit in under 5 minutes.

1. Detect Vulnerabilities (Fastest)

python3 -m src.main detect data/contracts/VulnerableBank.sol

This runs Slither + Aderyn + LLM analysis and outputs a table:

┌─────────────────────────────────────────┐
│       Vulnerabilities Found             │
├──────┬──────────────┬──────────┬────────┤
│ ID   │ Type         │ Severity │ Source │
├──────┼──────────────┼──────────┼────────┤
│ slither-0 │ reentrancy │ High     │ slither│
│ mimo-0    │ access_ctrl│ Critical │ mimo   │
└──────┴──────────────┴──────────┴────────┘

2. Full Audit (Detect + Patch + Verify)

python3 -m src.main audit data/contracts/VulnerableBank.sol --mode all

This runs the full 5-agent pipeline:

  1. Detect — Slither + Aderyn + LLM + RAG find vulnerabilities
  2. Patch — Architect designs strategy, Code Generator produces patches, Refiner iterates
  3. Verify — Validator runs Foundry tests to confirm fixes

Output is a markdown audit report with findings, patches, and verification status.

Limit Patches

# Patch only the 2 most severe vulnerabilities (default)
python3 -m src.main audit data/contracts/VulnerableBank.sol

# Patch all vulnerabilities
python3 -m src.main audit data/contracts/VulnerableBank.sol --max-patches -1

# Patch top 5
python3 -m src.main audit data/contracts/VulnerableBank.sol --max-patches 5

Save Report to File

python3 -m src.main audit data/contracts/VulnerableBank.sol -o report.json

3. Audit with On-Chain Attestation

python3 -m src.main audit data/contracts/VulnerableBank.sol \
  --attest \
  --contract-address 0xYourContract

This audits the contract AND posts the result to EAS Sepolia. See EAS Attestation Guide.

4. Multi-Expert Analysis

For deeper analysis using the forefy/.context multi-expert framework:

python3 -m src.main detect data/contracts/VulnerableBank.sol --multi-expert

This runs three parallel LLM "experts" — a systematic auditor, a fresh-perspective auditor, and a triager that validates findings.

Understanding the Output

Each vulnerability includes:

Field Meaning
id Unique identifier (e.g., slither-0, mimo-1)
type Vulnerability class (e.g., reentrancy, access_control)
severity critical / high / medium / low / informational
confidence Consensus score [0–1] — how many detector families agreed
source Detection engine (slither, aderyn, mimo, ba, ta, etc.)
verified Whether the Verificator confirmed it as a true positive

Next Steps

See Also