Skip to content

Commit 13e37fe

Browse files
SevenDataAISevenDataAI
committed
Initial RuMa Runtime
0 parents  commit 13e37fe

28 files changed

Lines changed: 3930 additions & 0 deletions

‎.gitattributes‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
* text=auto
2+
*.ps1 text eol=crlf

‎.github/workflows/ci.yml‎

Lines changed: 25 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,25 @@
1+
name: CI
2+
3+
on:
4+
push:
5+
branches:
6+
- main
7+
- "codex/**"
8+
pull_request:
9+
10+
jobs:
11+
verify:
12+
runs-on: ubuntu-latest
13+
14+
steps:
15+
- uses: actions/checkout@v4
16+
17+
- uses: actions/setup-node@v4
18+
with:
19+
node-version: 22
20+
cache: npm
21+
22+
- run: npm ci
23+
- run: npx playwright install --with-deps chromium
24+
- run: npm run build
25+
- run: npm run test

‎.gitignore‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,7 @@
1+
node_modules
2+
dist
3+
playwright-report
4+
test-results
5+
automation/logs
6+
automation/reports
7+
automation/runtime

‎LICENSE‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
MIT License
2+
3+
Copyright (c) 2026
4+
5+
Permission is hereby granted, free of charge, to any person obtaining a copy
6+
of this software and associated documentation files (the "Software"), to deal
7+
in the Software without restriction, including without limitation the rights
8+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9+
copies of the Software, and to permit persons to whom the Software is
10+
furnished to do so, subject to the following conditions:
11+
12+
The above copyright notice and this permission notice shall be included in all
13+
copies or substantial portions of the Software.
14+
15+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21+
SOFTWARE.

‎README.md‎

Lines changed: 103 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,103 @@
1+
# RuMa Runtime
2+
3+
[中文](./README.zh-CN.md) | English
4+
5+
![License](https://img.shields.io/badge/license-MIT-111111.svg)
6+
![Clients](https://img.shields.io/badge/clients-Codex%20%7C%20Claude%20Code%20%7C%20OpenClaw-1f6feb.svg)
7+
![Testing](https://img.shields.io/badge/tested-Playwright%20smoke%20coverage-0f766e.svg)
8+
9+
RuMa Runtime turns a prompt gallery into an agent operating runtime.
10+
11+
It combines:
12+
13+
- reusable operating modes such as `diagnose`, `recover`, `ship`, and `audit`
14+
- execution overlays such as `neutral`, `high-agency`, `hardline`, and `pua`
15+
- one-command installers for `Codex`, `Claude Code`, and `OpenClaw`
16+
- a browsable web surface for selecting modes and copying live prompt compositions
17+
- local QA and autopilot loops for repeated verification
18+
19+
The repo structure takes cues from two directions:
20+
21+
- the productized multi-client installation style of [`tanweai/pua`](https://github.com/tanweai/pua)
22+
- the content-library and reference-driven style of [`puaclaw/PUAClaw`](https://github.com/puaclaw/PUAClaw)
23+
24+
## Quick Start
25+
26+
```bash
27+
npm install
28+
npm run dev
29+
```
30+
31+
## Install Into Agents
32+
33+
```bash
34+
npm run install:codex
35+
npm run install:claude
36+
npm run install:openclaw
37+
npm run install:all
38+
```
39+
40+
Installed locations:
41+
42+
- Codex: `~/.codex/skills/ruma-runtime/SKILL.md` and `~/.codex/prompts/ruma-runtime.md`
43+
- Claude Code: `~/.claude/skills/ruma-runtime/SKILL.md`
44+
- OpenClaw: `~/.openclaw/skills/ruma-runtime/SKILL.md`
45+
46+
## Verify
47+
48+
```bash
49+
npm run build
50+
npm run test
51+
npm run check
52+
```
53+
54+
The Playwright suite covers:
55+
56+
- rendering every mode card
57+
- opening every modal
58+
- flavor and adapter switching
59+
- copy-to-clipboard flows
60+
- mobile layout coverage
61+
- installer copy behavior for all supported clients
62+
63+
## Autopilot
64+
65+
```bash
66+
npm run autopilot:register
67+
npm run autopilot:once
68+
npm run qa:loop
69+
```
70+
71+
The Windows autopilot loop runs build and smoke tests, writes reports to `automation/reports`, and can trigger one bounded backlog improvement when a local `codex` CLI is available.
72+
73+
## Repo Layout
74+
75+
```text
76+
adapters/
77+
claude/ruma-runtime/SKILL.md
78+
codex/ruma-runtime/SKILL.md
79+
openclaw/ruma-runtime/SKILL.md
80+
cursor/ruma-runtime.mdc
81+
commands/
82+
ruma-runtime.md
83+
references/
84+
failure-patterns.md
85+
runtime-playbook.md
86+
scripts/
87+
install-agent-skill.mjs
88+
install-codex-skill.mjs
89+
autopilot-once.ps1
90+
src/
91+
data/library.js
92+
tests/
93+
app.spec.js
94+
```
95+
96+
## References
97+
98+
- [`references/runtime-playbook.md`](./references/runtime-playbook.md)
99+
- [`references/failure-patterns.md`](./references/failure-patterns.md)
100+
101+
## License
102+
103+
[MIT](./LICENSE)

‎README.zh-CN.md‎

Lines changed: 103 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,103 @@
1+
# RuMa Runtime
2+
3+
中文 | [English](./README.md)
4+
5+
![License](https://img.shields.io/badge/license-MIT-111111.svg)
6+
![Clients](https://img.shields.io/badge/clients-Codex%20%7C%20Claude%20Code%20%7C%20OpenClaw-1f6feb.svg)
7+
![Testing](https://img.shields.io/badge/tested-Playwright%20smoke%20coverage-0f766e.svg)
8+
9+
RuMa Runtime 不是再做一个提示词仓库，而是把提示词、skill、执行协议和验证闭环压成一个可安装的 Agent Runtime。
10+
11+
它同时包含：
12+
13+
- `diagnose`、`recover`、`ship`、`audit` 这类运行模式
14+
- `neutral`、`high-agency`、`hardline`、`pua` 这类执行风格
15+
- `Codex / Claude Code / OpenClaw` 三端一键安装
16+
- 一个可浏览、可筛选、可复制的 Web 门面
17+
- 本地 QA 与 autopilot 循环
18+
19+
整体风格参考了两个方向：
20+
21+
- [`tanweai/pua`](https://github.com/tanweai/pua) 的多客户端安装与产品化目录
22+
- [`puaclaw/PUAClaw`](https://github.com/puaclaw/PUAClaw) 的内容库、参考资料和 GitHub 呈现方式
23+
24+
## 快速开始
25+
26+
```bash
27+
npm install
28+
npm run dev
29+
```
30+
31+
## 一键安装到 Agent
32+
33+
```bash
34+
npm run install:codex
35+
npm run install:claude
36+
npm run install:openclaw
37+
npm run install:all
38+
```
39+
40+
默认安装位置：
41+
42+
- Codex: `~/.codex/skills/ruma-runtime/SKILL.md` 和 `~/.codex/prompts/ruma-runtime.md`
43+
- Claude Code: `~/.claude/skills/ruma-runtime/SKILL.md`
44+
- OpenClaw: `~/.openclaw/skills/ruma-runtime/SKILL.md`
45+
46+
## 验证
47+
48+
```bash
49+
npm run build
50+
npm run test
51+
npm run check
52+
```
53+
54+
Playwright 冒烟覆盖包括：
55+
56+
- 全部模式卡片渲染
57+
- 全部 modal 打开
58+
- Flavor 和 Adapter 切换
59+
- 复制按钮行为
60+
- 移动端布局
61+
- 三端安装脚本复制结果
62+
63+
## 自治循环
64+
65+
```bash
66+
npm run autopilot:register
67+
npm run autopilot:once
68+
npm run qa:loop
69+
```
70+
71+
Windows 计划任务会按 5 分钟节奏巡检，写报告到 `automation/reports`，并在本机 `codex` CLI 可用时执行一次有边界的 backlog 自改。
72+
73+
## 仓库结构
74+
75+
```text
76+
adapters/
77+
claude/ruma-runtime/SKILL.md
78+
codex/ruma-runtime/SKILL.md
79+
openclaw/ruma-runtime/SKILL.md
80+
cursor/ruma-runtime.mdc
81+
commands/
82+
ruma-runtime.md
83+
references/
84+
failure-patterns.md
85+
runtime-playbook.md
86+
scripts/
87+
install-agent-skill.mjs
88+
install-codex-skill.mjs
89+
autopilot-once.ps1
90+
src/
91+
data/library.js
92+
tests/
93+
app.spec.js
94+
```
95+
96+
## 参考内容
97+
98+
- [`references/runtime-playbook.md`](./references/runtime-playbook.md)
99+
- [`references/failure-patterns.md`](./references/failure-patterns.md)
100+
101+
## 许可
102+
103+
[MIT](./LICENSE)

‎adapters/claude/ruma-runtime/SKILL.md‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,7 @@
1+
---
2+
name: ruma-runtime
3+
description: "Agent operating runtime for diagnose, recover, ship and audit workflows with neutral, high-agency, hardline and pua overlays."
4+
license: MIT
5+
---
6+
7+
Load the mode that best matches the current task and execute it with evidence-first discipline.

‎adapters/codex/ruma-runtime/SKILL.md‎

Lines changed: 28 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,28 @@
1+
---
2+
name: ruma-runtime
3+
description: "一个面向 Agent 的运行时 skill。把 diagnose、recover、ship、audit 等模式和 high-agency / hardline / pua 语气层打包成统一协议。适用于调试、实现、交付验证、部署审计和复杂多步任务。触发：卡壳、重复路线、准备甩锅、需要结构化输出、需要上线前自检。"
4+
license: MIT
5+
---
6+
7+
# RuMa Runtime
8+
9+
你不是来给用户堆提示词样板的。你要做的是在以下模式里选对一个，并带着证据交付：
10+
11+
1. `diagnose`：读失败信号、搜索、读原始材料、验证假设、给出本质不同方案
12+
2. `recover`：连续失败后先自救，再扩圈，而不是一直硬拧
13+
3. `ship`：未验证=未完成。build、test、curl、手动路径都要有证据
14+
4. `audit`：平台约束、依赖链、部署差异一次性审计，不打地鼠
15+
16+
## Flavor 选择
17+
18+
- `neutral`：默认，结构化清晰输出
19+
- `high-agency`：主动出击，扫描同类问题
20+
- `hardline`：高压执行，只认事实、动作和证据
21+
- `pua`：带压力的话术包装，但仍以动作和验证为主
22+
23+
## 基础规则
24+
25+
- 先做后问。问用户前必须附带你已完成的排查。
26+
- 同一路线第二次失败，必须换本质不同方案。
27+
- 声称完成前，必须先运行验证命令。
28+
- 修一个点后，主动检查同文件、同模块、同链路类似问题。

‎adapters/cursor/ruma-runtime.mdc‎

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,11 @@
1+
---
2+
description: RuMa Runtime for diagnose, recover, ship and audit workflows with evidence-first execution.
3+
alwaysApply: false
4+
---
5+
6+
# RuMa Runtime
7+
8+
- Diagnose before guessing.
9+
- Recover before grinding the same route.
10+
- Ship only with validation evidence.
11+
- Audit platform constraints end-to-end.

‎adapters/openclaw/ruma-runtime/SKILL.md‎

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
---
2+
name: ruma-runtime
3+
description: "OpenClaw runtime skill for diagnose, recover, ship and audit workflows with neutral, high-agency, hardline and pua overlays. Use when the agent is stuck, repeating the same fix, preparing to blame the environment, or needs a structured verification-first delivery loop."
4+
license: MIT
5+
---
6+
7+
# RuMa Runtime
8+
9+
Run the current task in one of four modes and keep the output evidence-first.
10+
11+
## Modes
12+
13+
1. `diagnose`: Read the failure signal, search the exact problem, inspect original context, verify assumptions, and propose materially different next steps.
14+
2. `recover`: Break stalled work into the smallest action that can recover new information.
15+
3. `ship`: Treat unverified work as incomplete. Run build, test, curl, or the real happy path before claiming success.
16+
4. `audit`: Inspect platform limits, dependency edges, environment differences, and rollout risks in one pass.
17+
18+
## Overlays
19+
20+
- `neutral`: Clear structure, minimal drama.
21+
- `high-agency`: Expand the blast radius check after the first fix.
22+
- `hardline`: No speculation without evidence. No status without proof.
23+
- `pua`: Add pressure language, but keep the actual mechanism grounded in actions and verification.
24+
25+
## Operating Rules
26+
27+
- Search, inspect, and execute before asking the user for missing context.
28+
- Switch to a materially different approach after the second failure on the same path.
29+
- Report with facts, actions, results, conclusion, and next step.
30+
- After fixing one issue, scan the same file, module, and flow for similar defects.

0 commit comments

Comments
 (0)