Chalk Labs
Need help deploying help?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/zxdxjtu-help
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/help
What's inside
- Skill name
help- Description
- Explain Harness plugin and available commands
- License
- MIT
- Instructions
- 127 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Harness — Spec-Driven Development Plugin Explain the following to the user: ## What is Harness? Harness 是一个自进化的 AI 辅助开发框架。你只需要告诉它做什么,它会自动走完 Spec → 测试 → 分解 → 执行 → 验证的全流程。 ## 你只需要记住一个命令 ``` /harness 实现用户登录功能 ``` Harness 会自动: 1. 了解你的项目(首次使用时) 2. 和你讨论需求、生成 Spec(需要你确认) 3. 生成测试合约(需要你确认) 4. 分解任务、按 Wave 并行执行(全自动) 5. 三阶段验证并输出证据(全自动) 6. 从失败中学习,下次做得更好(全自动) 整个过程你只需要参与 **2 次**:确认 Spec + 确认测试。 ## 用法 | 命令 | 作用 | |------|------| | `/harness 功能描述` | 从想法到交付,全流程 | | `/harness status` | 查看状态,接着干 | | `/harness init` | 仅初始化(了解项目和团队规范) | | `/harness evolve` | 手动看一下进化状态 | | `/harness` | 自动判断该做什么 | | `/cancel-sprint` | 紧急停止正在跑的 Sprint | ## 核心原则 1. **Spec → Test → Code** — 测试是人和 AI 的对齐合约 2. **原子任务** — 每个任务独立可测、可并行 3. **闭环验证** — 有证据才算完成 4. **自进化** — Sprint 失败自动沉淀为不变量,下次避免 5. **对抗分离** — 写代码的和审代码的不是同一个 Agent 6. **Pitfalls 记忆** — 模型不需要项目变迁历史,只需要"现状"和"哪些坑别踩" ## 自进化机制 **你不需要手动触发进化。** Harness 在以下时机自动学习: - **Sprint 完成时**:分析 JSONL 追踪日志,提取失败模式 - **任务失败时**:记录结构化失败原因 - **同一模式出现 3 次**:自动提升为不变量,注入后续 Sprint 进化结果保存在 `.harness/invariants.md` 和 `.harness/evolution-log.md`。 ## 高级用法 以下命令供了解内部机制的用户使用,普通使用不需要: | 命令 | 作用 | 何时需要 | |------|------|---------| | `/proposal` | 单独编写 Spec | 想精细控制需求阶段 | | `/tdd-align F001` | 单独生成测试 | 想单独审查测试 | | `/decompose F001` | 单独分解任务 | 想调整任务粒度 | | `/sprint F001` | 单独跑 Sprint | 想从中间阶段继续 | | `/verify F001` | 单独跑验证 | 想重新验证 | | `/adversarial-review F001` | 对抗式代码审查 | 想在合并前深度审查 | | `/baseline <url>` | 采集参照产品基线 | 克隆场景 | | `/evaluate F001` | 对标评估 | 克隆场景 | | `/eval-fix F001` | GAN 修复循环 | 克隆场景评分不够 | ## 项目状态目录 所有状态保存在 `.harness/`,可 gitignore 或 commit: ``` .harness/ ├── config.yaml # 项目配置(项目级,永久保留) ├── norms.md # 团队规范(项目级) ├── pitfalls.md # 踩坑记录(项目级,跨 feature 持久保留) ├── invariants.md # 学到的约束(项目级,自动增长) ├── specs/ # 功能规格(项目级) ├── evidence/ # 验证证据(项目级) ├── traces/ # JSONL 事件追踪(项目级) ├── skill-context/ # 上下文注入碎片(项目级) ├── evolution-log.md # 进化历史(项目级) ├── tasks.md # 任务 DAG(Feature 级,完成后归档) ├── progress.md # 进度日志(Feature 级,完成后归档) ├── archive/ # 已归档的 tasks/progress └── sprint-loop.md # Sprint 运行状态(Session 级,自动清理) ``` **建议 commit**: `pitfalls.md`、`invariants.md`、`specs/`、`config.yaml`、`norms.md` ## Pitfalls 机制 `.harness/pitfalls.md` 是轻量级的项目踩坑记录,与 invariants(重量级不变量)互补: | | pitfalls.md | invariants.md | |---|---|---| | **门槛** | 即时记录,无需验证 | 同一模式 3 次失败才提升 | | **格式** | 一行一条 | 结构化(规则+证据+检测方法) | | **维护** | 自动收集 + 手动编辑 | 自进化引擎自动管理 | | **本质** | 经验直觉 | 验证过的规律 | **自动流转**: `/proposal` 读取 → worker 读取 → wave 完成后收集 → 失败时提取 → `/verify` 收集 **手动编辑**: 随时可以直接编辑 `.harness/pitfalls.md`,格式:`- [模块]: 一句话描述` ## 文件生命周期 Feature 完成后,harness 自动清理: - `tasks.md`、`progress.md` → 归档到 `archive/` - `decisions/`、`attempts/`、`research/` → 删除 - `pitfalls.md`、`invariants.md`、`specs/`、`evidence/` → **永久保留** 这样下一个 feature 从干净状态开始,不会被陈旧 plan 误导。 ## 提示
From zxdxjtu/harness, MIT.
More from zxdxjtu/harness
- 0.466score
tdd-align
zxdxjtu
Generate three-layer tests from spec — tests are the alignment contract
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
evaluate
zxdxjtu
Adversarial evaluation: compare dev product against reference baseline using Playwright MCP
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
harness-init
zxdxjtu
Initialize Harness for your project — learn repo, team norms, generate customized framework
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
harness-status
zxdxjtu
Restore state, health check, determine current phase, suggest next step
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
proposal
zxdxjtu
Start a new feature proposal — guided spec + design generation
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
sdd-init
zxdxjtu
Initialize SDD harness for a project — scan codebase, generate config, full-spec, full-design
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
spec-review
zxdxjtu
Multi-person spec review with line-by-line comments and resolution tracking
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
steward
zxdxjtu
Module stewardship guardian — check code changes against AGENT.md boundaries and contracts
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
archive
zxdxjtu
Phase 3: merge spec/design deltas, code review, atomic commits, push, and trigger self-evolution
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
baseline
zxdxjtu
Capture baseline of a reference product using Playwright MCP for clone/replicate scenarios
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
entropy-clean
zxdxjtu
Post-wave entropy cleanup — remove debug artifacts, enforce style, ensure git hygiene
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
eval-fix
zxdxjtu
Adversarial fix loop: read eval report → fix gaps → re-evaluate → repeat until convergence
SkillDFreshness: D★ 8Claude Code +9
Signals
- Stars
- 8
- Forks
- 3
- Open issues
- 1
- Last commit (this folder)
- Apr 12, 2026 (6 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/zxdxjtu/harness
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from zxdxjtu
- 0.466score
eval-fix
zxdxjtu
Adversarial fix loop: read eval report → fix gaps → re-evaluate → repeat until convergence
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
evaluate
zxdxjtu
Adversarial evaluation: compare dev product against reference baseline using Playwright MCP
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
baseline
zxdxjtu
Capture baseline of a reference product using Playwright MCP for clone/replicate scenarios
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
entropy-clean
zxdxjtu
Post-wave entropy cleanup — remove debug artifacts, enforce style, ensure git hygiene
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
archive
zxdxjtu
Phase 3: merge spec/design deltas, code review, atomic commits, push, and trigger self-evolution
SkillDFreshness: D★ 8Claude Code +9 - 0.466score
harness-init
zxdxjtu
Initialize Harness for your project — learn repo, team norms, generate customized framework
SkillDFreshness: D★ 8Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get help running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
help skill. Agent Mart. https://agentmart.dev/skills/zxdxjtu-help (last updated 2026-10-24)