What makes Mev Engine notable?

1.4K downloads. Trust evidence available

How should Mev Engine be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Mev Engine

MEV Engine v8.0 ⚔️ — OpenClaw原生。MEV五层指导思想+交付约定+教训生命周期，全部基于OpenClaw内置能力，零自定义脚本。 Skill: Mev Engine Owner: meta-evo-creator Summary: MEV Engine v8.0 ⚔️ — OpenClaw原生。MEV五层指导思想+交付约定+教训生命周期，全部基于OpenClaw内置能力，零自定义脚本。 Tags: execution:2.8.0, framework:2.8.0, latest:8.0.0, methodology:2.8.0 Version history: v8.0.0 | 2026-05-18T06:07:36.310Z | user v8.0.0: MEV Engine OpenClaw原生化。删除6个冗余自定义脚本，全面改用OpenClaw原生能力。MEV是驾驶手册，OpenClaw是引擎。零自定义脚本。 v7.2.1 | 2026-05-14T15:09:23.946Z | user v7.2.1: Self-audit — Plugin Di

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.4k

Updated

Oct 10, 2026

Version

8.0.0

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.4K downloads reported by the source. Last updated 10/10/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 10, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 10, 2026
Adoption signal: 1.4K downloadsadoption · observed Oct 10, 2026
Latest release: 8.0.0release · observed May 18, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s178d1h38evdy216hm4xzkw55s867x8s:mev-engine

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-meta-evo-creator-mev-engine/snapshot"

Documentation

CLAWHUB

151,168 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: mev-engine
description: |
  MEV Engine v8.0 ⚔️ — OpenClaw原生。MEV五层指导思想+交付约定+教训生命周期，全部基于OpenClaw内置能力，零自定义脚本。
homepage: https://github.com/meta-evo-creator/mev-engine
version: 8.0.0
metadata:
  openclaw:
    emoji: ⚙️
    requires: {}
---

# MEV Engine v8.0 ⚔️

> **v8.0: 全部基于 OpenClaw 原生能力。MEV是驾驶手册，OpenClaw是引擎。零自定义脚本。**

## 定位

MEV Engine 不是代码框架，是**思考方法论 + 交付约定**。

- 🧠 **MEV五层** → 怎么思考一个问题
- 📋 **G0-G4门禁** → 怎么检查一个产出
- 📝 **交付约定** → Sign-off / UNSOURCED / 证据映射
- 🔄 **教训生命周期** → 学到的东西怎么不丢失

**引擎是 OpenClaw。** 所有执行都走 OpenClaw 原生能力（sessions / subagents / tools / skills / memory / delivery）。

---

## 架构

```
┌─────────────────────────────┐
│  OpenClaw（引擎）            │
│  sessions · subagents       │
│  tools · skills · memory    │
│  cron · delivery · heartbeat│
└────────────┬────────────────┘
             │
    ┌────────┼────────┐
    ↓        ↓        ↓
┌────────┐┌────────┐┌────────┐
│ SOUL.md││plugins/││memory/ │
│ 内核    ││ 技能    ││ 记忆    │
│ 不可变  ││ 按需加载││ 三层体系 │
└────────┘└────────┘└────────┘
```

---

## MEV 五层（思考框架，不是代码流水线）

> 内核只做Suit（入口适配）。Sense~Evolve是各插件设计内部流程时的参考框架。

### ① Suit — 入口适配

**不做的事：** 不再跑 `node mev-prefight.cjs`。

**做的事：**
- 读上下文：OpenClaw 已注入 Framework版本、当前时间、工具列表
- 判定Tier：L1快答 / L2标准 / L3深度
- 激活插件：`memory_search(PLUGIN-REGISTRY)` → 匹配 → `read` 加载插件

**OpenClaw实现：** `read` + `memory_search` + skills auto-activation

### ② Sense — 感知采集

**搜索降级链（OpenClaw原生）：**
```
tavily__tavily_search → web_fetch → babata-browser → 标注「不可达」
```

工具选择表见 SOUL.md。

### ③ Think — 分析加工

视任务复杂度，激活对应插件：
- 纪检法规 → `discipline-inspect`（4-Agent编排 + RAG）
- 医学研究 → `med-research`（Scout→Draft→Review）
- 深度调研 → `deep-research`（L3专用）

偏误检查、ACH、证据映射内置于各插件。

### ④ Optimize — 交付检查

**G0-G4 门禁（OpenClaw原生）：**

| 门禁 | 检查什么 | OpenClaw实现 |
|:-----|:--------|:-----------|
| G0 覆盖度 | 几个信源？几个维度？ | 手动统计（tavily result count + web_fetch URL数） |
| G1 结构 | 缺哪段？ | 对照模板检查 |
| G2 分析 | 偏误？遗漏？证据链？ | bias-check + evidence_map |
| G3 交付 | IMA上传？推送？ | `ima-skill` + `wecom_mcp` |
| G4 复盘 | 日志写了？教训沉淀了？ | `write(memory/YYYY-MM-DD.md)` + `edit(plugin LEARNED PATTERNS)` |

**交付格式要求：**
```
🔒 DELIVERY CHECK
[✅/❌] G0 Coverage: {n} sources / {n} dimensions
[✅/❌] G1 Structure: 完整 / 缺失{list}
[✅/❌] G2 Analysis: bias={PASS/修正} gap={无遗漏/已标记} evidence_map={n}/{total}
[✅/❌] G3 Delivery: IMA={kb_name} push={sent/failed}
[✅/❌] G4 Evolve: trace={written}
```

### ⑤ Evolve — 进化沉淀

**教训生命周期（OpenClaw原生）：**

```
遇到教训 → edit(plugin LEARNED PATTERNS 段)
         → edit(core-lessons.plugin.md 索引)
         
激活：插件被Dispatch激活 → 教训自动加载
退役：插件30天未触发 → 教训随之休眠
```

每日日志：`write(memory/YYYY-MM-DD.md)`
长期记忆提炼：`edit(MEMORY.md)`（每几日从每日日志提炼）

---

## Sign-off Protocol

> 来源：Anthropic Financial-Services → 纪检场景同构。AI Drafts, Humans Sign Off.

每个分析类产出必须带审批节点。详见各插件 agent 指令。

---

## 跳过规则

| 条件 | 跳过 |
|:-----|:-----|
| L1 简单查询 | G0-G4 + checkpoint |
| Cron 隔离 | **禁止子代理**，强制 G0-G4 |
| 无需 IMA | G3 IMA=N/A |

---

## 插件生命周期（OpenClaw原生）

```
scene/ (试用) → 触发≥3次 → active/ (常驻)
active/ → 30天未触发 → dormant/ (休眠)
dormant/ → 同类问题复现 →

README.md

# MEV Engine v7.0 ⚙️

> **Kernel + Plugin Architecture.** Minimal immutable core, context-activated plugins, auto-dormancy.

**Mission → Environment → Verification** — A five-layer task execution engine.
The core execution framework of Babata OS. Each layer: core question + verification criteria + exception paths.

---

## v7.0 Architecture

```
┌─────────────────────────────┐
│      Core Kernel (immutable) │
│  Identity · MEV Skeleton     │
│  Tool Table · Safety Rules   │
└──────────┬──────────────────┘
           │
   ┌───────┼───────┐
   ↓       ↓       ↓
 active   scene   dormant
(常加载)  (按需)   (休眠)
```

**Lifecycle:** scene(30d trial) → active(triggered ≥3x) → dormant(30d unused) → scene(reactivate)

---

## MEV Five Layers

```
① Suit    → Prepare & adapt (G0 preflight + boundary check)
② Sense   → Gather & collect (hypothesis explicit, ≥2 sources)
③ Think   → Analyze & falsify (bias check + method selection)
④ Optimize → Deliver (G0-G4 五层递进门禁)
⑤ Evolve  → Reflect (lessons + framework audit)
```

---

## Delivery Gates

```
🔒 DELIVERY CHECK
[✅/❌] G0 Coverage: {n} sources
[✅/❌] G1 Structure: 完整
[✅/❌] G2 Analysis: bias={} evidence_map={}
[✅/❌] G3 Delivery: IMA={} push={}
[✅/❌] G4 Evolve: trace={}
```

---

## Changelog

| Version | Date | Changes |
|:----|:----|------|
| v7.0.0 | 2026-05-13 | **Kernel+Plugin architecture.** Core immutable, capabilities as plugins, auto-dormancy lifecycle. MEV skeleton preserved, specific rules extracted to plugins. |
| v6.5.0 | 2026-05-12 | Trust-but-verify: unified preflight, Agent E verify, IMA fallback |

---

## Install

```bash
git clone https://github.com/meta-evo-creator/mev-engine.git
clawhub install mev-engine
```

## Dependencies

Zero external dependencies. Requires Python for time-awareness probe.

## License

MIT

_meta.json

{
  "ownerId": "kn7cy2z5e60pxd0830tar97xwx866ydt",
  "slug": "mev-engine",
  "version": "8.0.0",
  "publishedAt": 1779084456310
}

references/eval-loop.md

# Eval Loop for Self-Improvement

Use this reference when a repeated failure should become a tested operational improvement instead of only a logged lesson.

## Goal

Do not only ask "what did we learn?"
Also ask:
- what is the current baseline?
- what exact guardrail or rule changed?
- how will we measure whether it helped?
- should we keep or discard the change?

## Use this loop for
- repeated Mission Control wording failures
- missing receipts / missing proof chains
- deploy closeout failures
- stale operator-facing surfaces
- repeated handoff mistakes between agents
- recurring SOP/checklist changes

## 1. Define the target

State one concrete thing you want to improve.

Examples:
- Hunter summary should always include concrete links and details
- ClawLite deploy closeout should never stop at code-ready status
- Mission Control front-end should render source links from structured fields

## 2. Write 3-5 binary evals

Each eval must be yes/no.

Examples for summary quality:
- Does the summary include at least one artifact path or URL?
- Does the summary include evidence links when external proof matters?
- Does the summary include a detail block describing what actually changed?
- Does the summary include the next handoff or recovery action?
- Does the operator-facing surface actually render these fields?

Examples for deploy closeout:
- Is the deployed commit hash recorded?
- Is a deployment ref/URL recorded?
- Was the production page or sitemap actually verified?
- Was a structured receipt written?
- Is the final state classified with the correct deploy-state vocabulary?

## 3. Capture baseline

Before changing the rule/SOP/skill/checklist:
- record the current failure pattern
- record which evals currently fail
- treat this as the baseline state

## 4. Change only one thing

Good changes:
- one wording rule
- one new checklist item
- one schema field
- one render mapping
- one validation step

Bad changes:
- rewriting everything at once
- adding five new rules at once
- changing wording and schema and code together unless absolutely required

## 5. Re-check and classify

After the single change:
- run the same evals again
- note which checks improved
- decide:
  - KEEP
  - DISCARD
  - PARTIAL_KEEP

## 6. Promotion rule

Only promote broadly reusable changes after they pass the eval loop or after operator review confirms the change materially reduced the failure.

## Suggested experiment entry format

```md
## [EXP-YYYYMMDD-XXX] experiment

**Logged**: ISO-8601 timestamp
**Priority**: medium | high | critical
**Status**: baseline | testing | keep | discard | partial_keep
**Area**: workflow | tools | product | growth | security | infra | ops

### Target
What repeated problem is being improved

### Baseline
What was failing before the change

### Mutation
The single change introduced

### Binary Evals
- [ ] Eval 1
- [ ] Eval 2
- [ ] Eval 3

### Result
What improved / did not improve

### Keep or Discard
keep | discard | partial_keep

### Meta

references/promotion-guide.md

# Promotion Guide

Use promotion only when a learning is broadly reusable.

## Promote to AGENTS.md
When the learning changes execution workflow.
Examples:
- deploy ownership rules
- acceptance ownership rules
- escalation timing

## Promote to TOOLS.md
When the learning is an environment/tool routing rule.
Examples:
- use Tavily before Brave
- key locations in Keychain
- browser session attach rules

## Promote to SOUL.md
When the learning is a behavior/principle rule.
Examples:
- do not let no-assignment closeout replace required deliverables
- do not treat shallow checks as full acceptance

## Promote to Obsidian
When the learning should become reusable operator material, marketing proof, or an operations note outside transient chat.

By default, Obsidian-style exports go to the local safe fallback:
- `.learnings/exports/obsidian/`

If you want a real vault destination, set `OBSIDIAN_LEARNINGS_DIR` explicitly before running the promotion script.
Always confirm the printed target path first, or use `--dry-run`.

Example:
- `node scripts/promote-learning.mjs obsidian "Reusable learning" --dry-run`
- then rerun without `--dry-run` after confirming the path

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.