CASE 06 · IN PROGRESS

Grimoire

Exploits proven, not guessed.

2026

PROJECT INFOS

A robot auditor that hunts blockchain vulnerabilities and proves them with a runnable exploit.

Grimoire is a robot auditor for open-source blockchain bounty programs. It reads smart contracts, node clients and bridges, hunts for vulnerabilities, then proves them. Most “AI auditors” hand the whole job to a model and ship confident, wrong reports. Grimoire holds a harder line: nothing is called proven until the exploit has been run against a copy of the live chain, and anything short of that is reported as an assessment, with the evidence it rests on. It works through the code in parallel rather than in one pass, and it names the parts it could not reach instead of quietly leaving them out. What you get per finding is a disclosure you can send: a written report and a proof the recipient can run.

PRODUCT SPEC
PROVEN NOT REACHED
A vulnerability is only called proven once the exploit has been run

WHAT IT PROVES

  1. 01

    A vulnerability is only called proven once the exploit has been run

  2. 02

    The report names the code it could not reach, so you can see the gaps

  3. 03

    Each finding arrives ready to submit, with a proof the recipient can run

SCOPE

  • PRODUCT SPEC
  • AGENT ARCHITECTURE
  • EXPLOIT PROVING
  • TRIAGE CONSOLE
  • DISCLOSURE PIPELINE

STACK

  • PYTHON
  • REACT 19
  • VITE
  • SOLIDITY
  • MCP
NEXT PROJECT Intelectra