RugProof

Rugproof your code before someone else does.

🌐 Live site: omermaksutii.github.io/RugProof 📦 Latest: v1.0.0 — 45 commands · 23 agents · 45 skills · 13 MCP servers · tested, offline-first, with rule packs, a benchmark, non-EVM coverage, and post-deploy monitoring

A Claude Code plugin that turns your editor into a full-stack smart contract security auditor: vulnerability detection, working exploit PoCs, mainnet-fork simulation, invariant generation, gas profiling, soulbound on-chain audit certificates, and shareable audit cards — Solidity + Vyper across every major EVM chain.

Install

From this GitHub source (works today)

# In Claude Code:
/plugin marketplace add omermaksutii/RugProof
/plugin install rugproof@omermaksutii-RugProof
/rugproof-init                              # 60-second setup
/audit examples/VulnerableVault.sol         # try it on the bundled vulnerable demo

From the official Anthropic marketplace

Pending review. Once listed at claude.com/plugins:

/plugin install rugproof

Submit your own plugin at claude.ai/settings/plugins/submit or platform.claude.com/plugins/submit.

Local dev

git clone https://github.com/omermaksutii/RugProof
cd RugProof
make build       # MCP servers + scripts
make test        # forge tests + NFT tests + MCP smoke test
make audit-demo  # runs the bundled reentrancy exploit (passes)

Quickstart

# In any Solidity project:
/audit                          # full audit on the repo
/audit src/Vault.sol            # audit one file
/quick-scan                     # fast pre-commit-style scan
/audit-live 0xabc... --chain ethereum  # audit a deployed contract by fork
/exploit <finding-id>           # write a working Foundry PoC
/report                         # render Markdown / HTML / PDF / PNG card

What's in the box

45 slash commands

Audit: /audit /audit-deep /audit-strict /audit-changes /audit-live /audit-history /audit-deps /audit-multi-chain /quick-scan /rug-check /score /explain Output: /report /card /remediate Exploit: /exploit /exploit-chain /exploit-live Simulation: /simulate /replay-incident Tests / proofs: /test-gen /invariant /fuzz /coverage /symbolic /prover Analysis: /gas /upgrade-safety /verify-deploy /diff-audit /audit-diff /pre-deploy /monitor Tooling integration: /slither /mythril Workflow: /rugproof-init /dismiss /verify-finding /bounty /bounty-submit /demo Notifications: /notify-slack /notify-discord /tweet On-chain: /mint-cert (Berachain soulbound audit certificate)

23 specialist subagents

Functional: attacker · defender · exploit-poc-writer · invariant-writer · gas-optimizer · remediation-suggester · report-writer · assembly-auditor

Protocol-specific: amm-specialist · lending-specialist · staking-specialist · bridge-specialist · governance-specialist · yield-aggregator-specialist · nft-specialist

Hot-topic specialists (2025): aa-specialist (ERC-4337) · crosschain-messaging-specialist (LayerZero V2 / CCIP / Hyperlane / Wormhole / Axelar) · restaking-specialist (EigenLayer / Symbiotic / Karak) · intents-specialist (ERC-7683 / UniswapX / CoW)

Language / chain / economics: vyper-specialist (Vyper compiler-class bugs) · l2-sequencer-specialist (rollup finality, sequencer uptime, L1↔L2) · economic-rug-specialist (owner powers + 0–100 rugability score) · zk-verifier-specialist (proof-verifier correctness)

45 auto-invoked vulnerability skills

A detection-skill library that auto-activates when Claude sees matching code patterns. Covers the full CWE/SWC catalog plus modern DeFi-specific issues:

Classic (19): reentrancy (incl. read-only and cross-contract) · access control · oracle manipulation · flash-loan attacks · MEV/front-running · signature replay · storage layout · initialization · unchecked calls · DoS vectors · integer issues · delegatecall risks · tx-context misuse · token compatibility · approval issues · EIP-6780 self-destruct · inline assembly · pragma & address hygiene · centralization risk.

Hot-topic 2025 (8): ERC-4337 account abstraction · cross-chain messaging · Permit2 / EIP-2612 · ERC-1271 contract signatures · Diamond / EIP-2535 · restaking & EigenLayer AVS · ERC-7683 intents · ERC-4626 inflation/donation.

v0.2 additions (8): ve-lock governance · fee-on-transfer accounting · signature malleability · MEV / PBS · liquidation cascades · oracle redundancy failure · cross-contract state inconsistency · ZK verifier bugs.

AI-quality meta-skills (4): confidence scoring · multi-pass self-critique (/audit-strict) · known-good reference comparison · false-positive feedback loop (/dismiss + .rugproof.yml ignore: + inline // rugproof-ignore markers).

DX meta-skills (2): caching-and-incremental (skip unchanged files) · progress-and-streaming (live status during long audits).

13 MCP servers

Chain I/O, test runners, history databases, static-analyzer + fuzz runners — block-explorer (Etherscan v2 multichain, EIP-1967/1822 proxy resolution) · forge-runner · hardhat-runner · anvil · tenderly · c4-history · sherlock-history · gas-tracker · token-metadata (+ GoPlus safety) · slither-runner · mythril-runner · fuzz-runner (Echidna/Medusa/Halmos). Every server degrades gracefully to labeled mock data offline, so the plugin works with zero configuration.

4 hooks

Pre-commit quick-scan · pre-push full audit · pre-deploy build-artifact check · post-test coverage analyzer. (Pre-commit and pre-push live as git hook scripts; pre-deploy and post-test wire into Claude Code's Bash hook events via plugin.json.)

5 bundled vulnerable demo contracts

examples/ ships a "wow on first install" set: classic reentrancy vault, spot-oracle lending market, flash-loan governance, ERC-4626 inflation attack, replay-able bridge. Run /demo after install to see Rugproof in action without bringing your own code.

Configuration

Create a .rugproof.yml at the repo root. See .rugproof.yml.example.

severity_threshold: high       # block CI on this severity or worse
chains: [ethereum, berachain, arbitrum, base]
include: ["src/**/*.sol"]
exclude: ["test/**", "script/**"]
ignore:
  - id: REENT-001
    reason: "Single-trusted-caller, reviewed 2025-Q4"

Inline suppression in code:

// rugproof-ignore: REENT-001 — single trusted caller, reviewed 2025-Q4
function trustedOp() external onlyOwner { ... }

Supported

Chains: Ethereum · Berachain · Arbitrum · Base · Optimism · Polygon · BSC · Linea · zkSync · Scroll

Languages: Solidity (full) · Vyper (specialist + skill) · Stylus/Rust · CosmWasm · Solana-Anchor (dedicated skills for top vuln classes) · Yul / inline assembly (specialist)

Output formats

/report produces Markdown, HTML, PDF, JSON, and a shareable PNG audit card. /card produces just the PNG — built for social sharing.

Live site

📖 Docs + sample reports + gallery: omermaksutii.github.io/RugProof

The site is built from the docs/ folder via the GitHub Pages workflow at .github/workflows/pages.yml — every push to main that touches docs/ redeploys.

Repository layout

RugProof/
├── .claude-plugin/plugin.json        # marketplace manifest + MCP/hook wiring
├── .github/                          # workflows (pages, pr-audit, release) + action + templates
├── commands/                         # 45 slash commands (real prompts)
├── agents/                           # 23 subagents
├── skills/                           # 45 auto-invoked detection skills
├── mcp/                              # 13 MCP servers (TypeScript)
├── scripts/                          # render-card · render-report · md-to-html · telemetry · …
├── hooks/                            # 4 hooks
├── templates/                        # report.md.hbs · report.html.hbs · audit-card.svg.hbs
├── examples/                         # 5 vulnerable demo contracts
├── nft/                              # AuditCertificate.sol (soulbound) + tests
├── samples/                          # rendered sample reports + cards
├── docs/                             # GitHub Pages site (auto-deployed)
├── rules/                            # community rule pack template
├── test/                             # Foundry exploit tests
├── lib/                              # forge-std, openzeppelin-contracts (gitmodules)
├── foundry.toml · remappings.txt
├── Makefile · LICENSE · SECURITY.md · CONTRIBUTING.md · CODE_OF_CONDUCT.md · CHANGELOG.md
├── .rugproof.yml.example · .rugproofignore.example
└── README.md

Build

One-liner via Makefile:

make build       # builds all 13 MCP servers + scripts
make test        # forge tests + NFT tests + MCP smoke test
make audit-demo  # runs the bundled reentrancy exploit PoC (passes)
make sample-cards   # rerender PNG audit cards from samples/
make sample-html    # rerender HTML pages from samples/

Or manually:

cd mcp && npm install && npm run build
cd ../scripts && npm install && npm run build
forge install                   # if you want to refresh git submodules
forge test -vv                  # exploit PoC against VulnerableVault passes
node scripts/dist/test-mcp.js   # MCP smoke test (13/13 servers should pass)

The plugin.json points to mcp/<name>-mcp/dist/index.js for each server. Most return mock data when API keys / external tools are absent so the plugin works out-of-the-box even without a fully configured environment.

Roadmap

v0.1 shipped the full skeleton; the 0.2 → 1.0 arc hardened it into a tested, real-integration, launch-ready tool.

Version	Scope	Status
0.1.0	Full skeleton: every command · agent · skill · MCP · hook · template · demo · NFT cert · Action · Pages site	✅
0.2.0	Bug-fixes + real test suites + CI gates + Etherscan v2 / GoPlus / Slither / Mythril MCPs + 4 specialists / 8 skills / 5 commands + source-generated docs	✅
0.3.0	EIP-1967/1822 proxy resolver · live-analyzer CI · dependency maintenance	✅
0.4.0	Rule-pack validator + 2 real packs · accuracy benchmark harness	✅
0.5.0	fuzz-runner MCP (Echidna/Medusa/Halmos) · non-EVM skills (Stylus/CosmWasm/Solana)	✅
0.6.0	Monitoring MCP + /monitor · Immunefi bounty formatter	✅
0.7.0	On-chain certificate deploy-readiness (Berachain runbook + targets)	✅
0.8.0	Source-driven public audit gallery	✅
0.9.0	Incremental cache · MCP-boundary input safety · _shared tests	✅
1.0.0	Launch: frozen findings schema + semver commitment + owner-gated handoff (LAUNCH.md)	✅

Shipping to users is the only remaining work, and it's external (marketplace submission, mainnet cert deploy, API keys, hosting) — see LAUNCH.md.

License

MIT