What makes Coding Lead notable?

3.2K downloads. Trust evidence available

How should Coding Lead 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

Coding Lead

Coding execution skill for any implementation-focused agent. Current production path is claude-only with simple tasks direct, medium tasks preferring ACP run...

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

3.2k

Updated

Oct 9, 2026

Version

3.0.6

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 3.2K downloads reported by the source. Last updated 10/9/2026.

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 9, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 9, 2026
Adoption signal: 3.2K downloadsadoption · observed Oct 9, 2026
Latest release: 3.0.6release · observed Mar 30, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17dg3qthhjbatdnrgq35n8jrh83j4ba:coding-lead

Install using `clawhub skill install s17dg3qthhjbatdnrgq35n8jrh83j4ba:coding-lead` in an isolated environment before connecting it to live workloads.
No published capability contract is available yet, so validate auth and request/response behavior manually.
Review the upstream CLAWHUB listing at https://clawhub.ai/beyound87/coding-lead before using production credentials.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-beyound87-coding-lead/snapshot"

Documentation

CLAWHUB

152,227 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: coding-lead
description: Coding execution skill for any implementation-focused agent. Current production path is claude-only with simple tasks direct, medium tasks preferring ACP run or direct acpx, and complex tasks handled via existing agent continuity plus context files instead of ACP session persistence. Integrates with qmd and smart-agent-memory when available.
---

# Coding Lead

> This is a coding execution skill for any agent that owns implementation work.
> It defines **how coding work runs**, not **who should own the task**.
> In multi-agent teams, routing may be handled elsewhere; in single-agent use, this skill still works directly.

Route by complexity. Current production path is claude-only. Do not depend on ACP session persistence in IM threads; use direct execution, direct acpx, and existing implementation session continuity instead.

## Task Classification

| Level | Criteria | Action |
|-------|----------|--------|
| **Simple** | Single file, <60 lines, clear local scope | Direct: read/write/edit/exec |
| **Medium** | 2-5 files, clear scope, likely follow-up questions | Prefer Claude ACP `mode:"run"` or direct acpx → fallback direct |
| **Complex** | Multi-module, architecture, needs continuity | Use existing implementation session continuity + context files + direct acpx/direct execution |

When in doubt, go one level up.

### Practical default rule
- **Simple**: stay in the current session; do not open ACP unless clearly beneficial
- **Medium**: prefer Claude ACP one-shot (`run`) when available; otherwise direct acpx
- **Complex**: preserve continuity through the existing implementation session, on-disk context files, and serial follow-ups; do not make IM-bound ACP `session` the default path
- **ACP unavailable**: medium/complex fall back to direct acpx or direct execution; simple tasks were already direct by default
- **Never block on ACP availability**: ACP is an accelerator, not a hard dependency

### Multi-agent dispatch rule (when dispatching to other agents)
- Dispatch to **ACP-configured agents** (agents with `runtime.type: "acp"` in openclaw.json): use `sessions_spawn(runtime="acp")`, may include `streamTo="parent"`
- Dispatch to **standard agents** (all others): use `sessions_spawn(runtime="subagent")`, **never include `streamTo`**
- Results from subagent spawns arrive via auto-announce, not polling
- How to tell which is which: check the target agent's config in openclaw.json — if it has `runtime.type: "acp"`, use ACP; otherwise use subagent

## Tech Stack (New Projects)

| Layer | Preferred | Fallback |
|-------|-----------|----------|
| Backend | PHP (Laravel/ThinkPHP) | Python |
| Frontend | Vue.js | React |
| Mobile | Flutter | UniApp-X |
| CSS | Tailwind | - |
| DB | MySQL | PostgreSQL |

Existing projects: follow current stack. New: propose first, wait for confirmation.

## Tool Detection & Fallback

All tools are **optional**. Detect once per session:

README.md

# coding-lead

## 这是什么
`coding-lead` 是一个通用编码执行技能,适用于任何承担实现职责的 agent。它负责定义**怎么做代码工作**,不负责决定**该由谁接任务**。

## 使用定位
- 单装可用:单 agent 直接使用
- 多 agent 可增强:团队里由路由/派工层决定谁来调用它
- 不依赖 team-builder 来源,只依赖是否存在编码执行任务

## 适用范围
适用于:
- 单项目开发
- 模块级实现与修复
- 需要连续上下文的代码链
- 需要在 ACP / direct acpx / direct execution 之间做稳定切换的开发任务

不适用于:
- 团队级派工与角色裁决
- 跨角色组织治理
- 非编码类任务规范

## 核心能力
### 1. 按复杂度路由执行方式
- **Simple**:当前会话内直接 `read/write/edit/exec`
- **Medium**:优先 Claude ACP `run`
- **Complex**:使用现有实现会话连续性 + context 文件 + 串行推进

### 2. ACP 检测与回退
每个 session 只检测一次,按顺序:
1. `sessions_spawn(runtime:"acp")`
2. `acpx claude exec`
3. direct execution

若当前模式中途失效,重新检测。

### 3. Context 文件治理
- active 上限:**60**
- 生命周期总窗口:**100**(active + archive)
- 推荐文件名:`<project>/.openclaw/context-<task-slug>.md`
- 同一代码链尽量复用同一个 context 文件

### 4. 精简 Prompt 策略
ACP prompt 只保留:
- 项目路径 / stack
- context 文件路径
- task
- acceptance criteria

项目级规则让 Claude Code 自己从 `CLAUDE.md` / `.cursorrules` / docs 中读取,不在 prompt 里重复灌入。

### 5. 完成前强制校验
宣布完成前必须再次核对:
- 任务目标
- 验收标准
- 显式禁区 / no-go constraints
- 是否存在无关改动

## 初始化
本技能**无需单独初始化**。

使用前需要满足:
- OpenClaw 已安装
- 当前承担编码执行的 agent 能读取本技能
- 如需 ACP:环境里存在可用的 `sessions_spawn(runtime:"acp")` 或 `acpx`

## 使用方法
### 方式一:作为自动执行规则使用
当会话加载本技能后,按技能内规则自动决定:
- 任务分类
- ACP 检测
- context 文件治理
- fallback 路径
- verify 与记录

### 方式二:按技能约定手动执行
典型流程:
1. 判断任务是 simple / medium / complex
2. 检测 ACP 是否可用
3. 如为 medium/complex,先写 context 文件
4. 用最小 prompt 发起 Claude ACP `run` 或 direct acpx
5. 完成后 verify + record + 清理 context

## 配置说明
本技能本身没有独立配置文件,但依赖以下外部条件:

### 1. ACP 运行环境
- `sessions_spawn(runtime:"acp", agentId:"claude")`
- 或 `acpx claude exec`

### 2. 团队标准文件(可选但推荐)
直接执行时,优先读取:
- `shared/knowledge/tech-standards.md`

### 3. 项目规则文件
ACP 执行时,项目自己的以下文件由 Claude Code 负责读取:
- `CLAUDE.md`
- `.cursorrules`
- 项目 docs

## 工作模式
### Simple
- 直接读文件
- 小范围修改
- 本地验证
- 记录变更

### Medium
- 建立 context 文件
- 用 Claude ACP `run` 执行
- 测试 / linter / acceptance 校验
- 记录经验

### Complex
- 延续现有实现会话
- 使用磁盘 context 文件保留上下文
- 串行推进同一代码链
- 仅在边界明确时并行,且总工作单元上限 5

## 边界与限制
- 不负责团队派工,只负责编码执行方式
- 不负责替代项目级规则系统
- 不建议在 `~/.openclaw/` 下直接 spawn 编码代理
- 同一代码链不要多工作单元并发写同一片文件

## 相关文件
- `SKILL.md`
- `references/prompt-templates.md`
- `references/complex-tasks.md`

## 最近一次修改（中文）
- **2026-03-29 / 64ff87c**
- 变更摘要：
  - context 活跃上限 **60**，生命周期总窗口 **100**
  - 强化完成前 verify 规则
  - 强化写入 / spawn 前确认 `cwd` 要求
  - 明确当前正式 ACP 编码路径为 **Claude-only**
  - complex 主路径为：现有实现会话连续性 + context 文件 + 串行推进

_meta.json

{
  "ownerId": "kn761rne5jajzfg0d8h6nge4hh8220cq",
  "slug": "coding-lead",
  "version": "3.0.6",
  "publishedAt": 1774831566153
}

references/complex-tasks.md

# Complex Tasks — Roles, QA Isolation, Parallel Strategies

> **Only read this file for Complex-level tasks.** Simple and Medium tasks don't need any of this.

## Complex Task Flow

```
RESEARCH → PLAN → EXECUTE → REVIEW → QA → FIX → RECORD
```

Skip steps that don't apply. Not every complex task needs all steps.

## Coding Roles

**Don't micromanage Claude Code's internal roles.** If Claude Code has its own agent system (e.g., oh-my-claudecode, project-level `.claude/agents/`), it handles role routing internally — architect, reviewer, QA, etc. Just give it the task.

OpenClaw's job for complex tasks:

```
1. RESEARCH (OpenClaw agent reads code, searches memory + qmd)
2. PLAN (agent designs approach, gets confirmation if needed)
3. EXECUTE (spawn ACP — one prompt, Claude Code handles internal delegation)
4. VERIFY (check acceptance criteria after completion)
5. RECORD (log to memory)
```

### When to spawn multiple ACP sessions (max 2)

Only when tasks are **truly independent** (different projects or zero file overlap):
```
Session 1: "Build favorites backend API..." (cwd: project-a)
Session 2: "Fix CSS layout bug..." (cwd: project-b)
```

For complex tasks within ONE project, prefer a single ACP session — Claude Code's internal agents parallelize better than multiple external spawns.

## QA Isolation

If Claude Code has internal QA agents (e.g., OMC test-engineer, qa-tester), it handles test isolation internally.

If you need to ensure isolation explicitly, mention it in the ACP prompt:
> "Write tests based on requirements and interfaces only, not implementation. Use a separate agent/session for tests."

The principle: tests should verify the contract, not mirror the code.

## RESEARCH Phase (No Changes)

Spawn in session mode, investigation only:

```
Investigate <project-dir> for <task description>.

DO NOT make any changes. Only report:
1. Files that need changes
2. Dependencies and call chains
3. Reusable existing code
4. Risks and edge cases
5. Test coverage status

Use qmd/grep/find to explore. Read .openclaw/context-xxx.md for project context.
```

## Agent Routing

Route via `agentId` in `sessions_spawn`. Fallback: preferred → alternate → direct execution.
Claude Code handles its own internal agent selection (OMC, project agents, etc.) — don't duplicate.

## Verification

Verification is mandatory before declaring success. Validate against the stated task, acceptance criteria, explicit constraints, and obvious regressions. If any of these remain unclear, report the gap instead of claiming completion.


| Level | What OpenClaw checks |
|-------|---------------------|
| Simple | No formal review |
| Medium | Success reported? Tests pass? |
| Complex | Acceptance criteria met? Tests pass? No unrelated changes? |

Code review itself is Claude Code's job (it has reviewers internally).

## Definition of Done

### Medium Tasks
- [ ] ACP reported success (or fallback completed)
- [ ] Linter passed (if available)
- [ ] Existing tests pass
- [ ] N

references/prompt-templates.md

# Prompt Templates & Examples

## Context File Template

Write to `<project>/.openclaw/context-<task-id>.md` before spawning:

```markdown
# Task Context: <task-id>

## Project
- Path: /path/to/project
- Stack: Laravel 10 + React 18 + TypeScript + Inertia.js
- Architecture: Action-based (not Service layer)
- Note: Project has CLAUDE.md with full standards — don't repeat here

## Historical Context
- [From memory recall] 2026-02-20: Switched to Action pattern
- [From memory lessons] PlatformLink has 6 types — use enum not magic strings
- [Known pitfall] common/ module auth: reuse CommonAuthTrait

## Long-term Knowledge Boundary
- Preserve only durable decisions, reusable lessons, and stable architecture facts
- Do not promote transient debugging noise, temporary paths, or one-off guesses

## Relevant Code (from qmd)
- app/Actions/UserAction.php — example of Action pattern
- app/Models/User.php — existing user model
- database/migrations/ — migration naming convention

## Related Lessons
- "Including DB schema upfront helps ACP get it right first try"
- "Always include existing test patterns as reference"
```

## Medium Task: Lean Prompt

```
Read .openclaw/context-feat-favorites.md for full project context.

## Task
Add a "favorites" feature: users can favorite items.
Need: migration, model, action, controller, API resource, React component.

## Acceptance Criteria
- [ ] Migration creates user_favorites pivot table
- [ ] Action handles add/remove/check
- [ ] Controller under 20 lines, delegates to Action
- [ ] React component with optimistic update
- [ ] Uses existing auth from common/
- [ ] Existing tests pass

Before finishing:
1. Run linter if available
2. Run tests if available
3. Include results in output

When finished: openclaw system event --text "Done: favorites feature" --mode now
```

**Note how small this is (~20 lines) vs the old approach (~60+ lines with embedded context.** The context file does the heavy lifting.

## Complex Task: Research Phase

```
Read .openclaw/context-refactor-collectors.md for project context.

Investigate app/Services/Collectors/ for refactoring.

DO NOT make changes. Only report:
1. All collector files and their responsibilities
2. How Aggregator calls each collector
3. Error handling and retry patterns
4. Code duplication between collectors
5. Test coverage

Use grep/find to explore. Report findings in structured format.
```

## Complex Task: Execute Phase (after plan confirmed)

```
Read .openclaw/context-refactor-collectors.md for project context.

## Confirmed Plan
- Extract shared HttpClientTrait from 3 collectors
- Do NOT change the collector interface
- Only refactor internal implementation

## Task
1. Create app/Traits/HttpClientTrait.php with shared retry/timeout logic
2. Refactor PaymentCollector, InventoryCollector, ShippingCollector to use trait
3. Keep Aggregator unchanged
4. Run existing tests

## Acceptance Criteria
- [ ] HttpClientTrait with configurable retry (default 3)
- [ ] All 3 collecto

Github ReposUpdated 3h 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.