Cleo (@cleo_x7f2)

Cleo is an OpenClaw-style autonomous X (Twitter) agent designed to operate in an adversarial public environment.

• X account: @cleo_x7f2
• Public Base wallet: 0xf99b68cc16b2d470e78d5d3a73ec1396c1786eeb
• Wallet explorer: https://basescan.org/address/0xf99b68cc16b2d470e78d5d3a73ec1396c1786eeb
• Security layer: Translcideus (https://translcideus.com)
• Creator: https://github.com/Cormacwren

What this repo gives you

A production-minded starter that actually runs:

• ✅ X posting + replying (Twitter API v2)
• ✅ “Challenge mode”: aggressively sanitizes/filters inputs to reduce jailbreak surface
• ✅ Base (L2) wallet module (ethers v6) for balances + transaction simulation hooks
• ✅ Pluggable “Translcideus” client (HTTP) — stubbed to integrate with your service
• ✅ Threat model + security checklist
• ✅ Docker + dotenv

Important: This repo is safe-by-default. It does not auto-transfer funds. You must explicitly enable any on-chain actions and implement your own approval rules.

---

Quick start

1) Requirements

• Node.js 20+
• A Twitter/X developer app + API keys
• (Optional) OpenAI-compatible model endpoint key
• (Optional) Base RPC (Alchemy/Infura/QuickNode, or public RPC)
• (Optional) Translcideus API endpoint + key

2) Install

npm install
cp .env.example .env

3) Configure .env

Fill in:

• X_* keys
• MODEL_* keys (if you want LLM responses)
• BASE_RPC_URL + AGENT_PRIVATE_KEY (if you want on-chain reads/writes)
• TRANSLCIDEUS_* (if you want security verdicts)

4) Run

Dev:

npm run dev

Prod:

npm run build
npm start

---

How it works (high-level)

• src/index.ts starts two loops:

  1. A tweet scheduler (periodic “tower log” style posts)
  2. A mentions/replies listener that:
     • pulls new mentions
     • runs them through security filtering
     • generates a response (rule-based or model-based)
     • posts the reply

• The wallet module is isolated and gated.

  • It can fetch balances and recent txs
  • It exposes disabled-by-default hooks for on-chain actions

• docs/THREAT_MODEL.md documents attack surfaces + mitigations.

---

Project structure

src/
  agent/cleo.ts              # agent brain (prompts + response policy)
  twitter/client.ts          # Twitter API v2 wrapper
  security/translcideus.ts   # optional security verdicts
  policies/policy.ts         # prompt-injection + jailbreak filters
  wallet/baseWallet.ts       # Base chain utilities (ethers v6)
  util/logger.ts             # structured logs
docs/
  THREAT_MODEL.md
  SECURITY.md

---

License

MIT — see LICENSE