REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Repeatable agents-plus-code workflows, packaged as one skill, stamped into any repo. Deterministic Python owns the graph; coding agents are bounded nodes inside it.
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/disler-super-simple-software-factory)在同一工作中把 AI 连到服务的 MCP
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Codebase intelligence for TypeScript and JavaScript, exposed as structured MCP analysis tools.
MCP server for terminal commands, file operations, and process management
同类工作,排序与 Skill 列表相同
| 本项 super-simple-software-factory | rea | andrej-karpathy-skills | graphify | MoneyPrinterTurbo | |
|---|---|---|---|---|---|
| 作者 | disler | morluto | multica-ai | Graphify-Labs | harry0703 |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 1,056 | 467,235 | 234,125 | 149,873 | 144,332 |
| 最近代码更新 | 2026年8月4日 | 2026年11月18日 | 2026年4月20日 | 2026年11月18日 | 2026年11月18日 |
| 许可证 | MIT | MIT | – | Apache-2.0 | MIT |
disler 提供的 1 个 Skill(指令集),AI 读取后照做 · 不是 MCP server
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
Claude CodeCursor
依据开发者 README 和 repo 中的文件夹
GitHub account disler (person), 12 years old
★ 1,056 stars · code last changed 2026年8月4日
46 scripts that can run on a computer: apps/visualizer/server/db.ts, apps/visualizer/server/index.ts, apps/visualizer/shared/types.ts, apps/visualizer/src/lib/api.ts and 42 more
Could not check every script (some are too large); read them first
No hooks or writes to machine settings found
Checked automatically from the repo files at commit de31374 (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
Uploading to claude.ai would fail: the description is over 200 characters (357). Use "Let the AI install it" instead
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /sssf,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/disler/super-simple-software-factory.git /tmp/popnx-skill-disler-super-simple-software-factory-de31374 && mkdir -p ~/.claude/skills && cp -R '/tmp/popnx-skill-disler-super-simple-software-factory-de31374/.claude/skills/sssf' ~/.claude/skills/sssfChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
把 sssf 文件夹放到 ~/.agents/skills/sssf
~ 是你的主文件夹($HOME)
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/disler/super-simple-software-factory.git /tmp/popnx-skill-disler-super-simple-software-factory-de31374 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-disler-super-simple-software-factory-de31374/.claude/skills/sssf' ~/.agents/skills/sssfChecked against official docs 2026年11月15日 · Source: learn.chatgpt.com
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/disler/super-simple-software-factory.git /tmp/popnx-skill-disler-super-simple-software-factory-de31374 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-disler-super-simple-software-factory-de31374/.claude/skills/sssf' ~/.agents/skills/sssf复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 sssf
Checked against official docs 2026年11月12日 · Source: cursor.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/disler/super-simple-software-factory.git /tmp/popnx-skill-disler-super-simple-software-factory-de31374 && mkdir -p ~/.cursor/skills && cp -R '/tmp/popnx-skill-disler-super-simple-software-factory-de31374/.claude/skills/sssf' ~/.cursor/skills/sssf适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
✓ 命令已检查 2026年11月18日
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
7,220 个字符 · 在 GitHub 打开完整文件
---
name: sssf
description: Super Simple Software Factory — deploy and operate repeatable agents+code workflows (ADWs) in any codebase. Use when the user says /sssf install, wants to create/run/update an ADW, manage the agent roster in sssf.config.yaml, or observe running agent workflows. Keywords - sssf, software factory, ADW, AI developer workflow, agent pipeline, install factory.
argument-hint: "[install | create adw | run adw | update config | ...]"
---
# Super Simple Software Factory (SSSF)
Reusable combination of **agents plus code**: deterministic Python ADW scripts own sequencing, retries, and acceptance; coding agents (Pi in v1) work inside bounded phases; typed JSON envelopes carry context between them; everything streams into SQLite for the polled visualizer. Agent proposes, code disposes.
## Startup
Three steps. Then stop.
1. Read [cookbooks/sssf_overview.md](cookbooks/sssf_overview.md) — the system map.
2. `ls adws/adw_*.py` and read each file's `Phases:` docstring line.
3. Print the ADWs as a table — name, the chain, one line on when to reach for it — and **wait for the engineer's request.**
```
| ADW | Chain | Use when |
|---|---|---|
| adw_scout | engineer → scout | read-only recon; nothing changes |
| adw_simple_sdlc | plan → build → test → review → document, 3 commits | the work is real and its shape is not obvious |
```
**Nothing else.** No trace-db queries, no reading the config or the ADW scripts' bodies, no repo inventory, no last-runs summary, no diagnosing an old failure, no "current state" dashboard. None of it was asked for, and it is not free:
- **Volunteered state is guessed state.** An orchestrator that improvised a status board queried a `runs` table and a `payload` column — neither exists (`sessions`, `payload_json`). The spec that would have said so is `references/observability.md`, one lazy read away. Probing to look prepared is how you end up confidently wrong in your first message.
- **It spends the context the real task needs**, before you know what the task is.
- **It is stale on arrival.** State printed before the request describes a system that the very next run changes.
Everything else — the db schema, the roster, the handoff contract — is lazy-loaded through the routing table below, when a request actually calls for it. Reading it early defeats the mechanism.
Two exceptions, both narrow: if the engineer's first message already contains a request, skip the waiting and route it; and if the factory is plainly not installed (no `adws/`, no config), say that in one line instead of the table.
## Orchestrator rules
You run the system, observe the system, and help the user interact with it. **You do no ADW work yourself:**
- Never implement, plan, or test in an agent's place — launch the ADW and watch it.
- Never edit files inside `adws/adw_data/sessions/` — that is the run record.
- Observe by querying `adws/adw_data/sssf.db` (WAL — reads never block writers) **when observing is the task**. This is a capability, not a startup step: query it to follow a run you launched or one the engineer asked about, never to volunteer a status report nobody requested.
- Report phase status plainly: name, owner, status, error if any.
## Request routing (lazy-load the cookbook, then follow it)
| Request | Cookbook |
|---|---|
| `/sssf install`, set up the factory in this repo | [cookbooks/install.md](cookbooks/install.md) |
| create a new ADW / workflow | [cookbooks/create_adw.md](cookbooks/create_adw.md) |
| modify an existing ADW chain | [cookbooks/update_adw.md](cookbooks/update_adw.md) |
| create the config / agent roster | [cookbooks/create_config.md](cookbooks/create_config.md) |
| add or retune an agent (model, thinking, tools, prompts) | [cookbooks/update_config.md](cookbooks/update_config.md) |
| extend adw_modules with new low-level logic | [cookbooks/update_modules.md](cookbooks/update_modules.md) |
| run / monitor an ADW | [cookbooks/how_to_prompt_for_the_eng.md](cookbooks/how_to_prompt_for_the_eng.md) **first**, then [cookbooks/run_adw.md](cookbooks/run_adw.md) |
| turn a request into an ADW prompt | [cookbooks/how_to_prompt_for_the_eng.md](cookbooks/how_to_prompt_for_the_eng.md) |
Deep specs, when needed: [references/config.md](references/config.md) · [references/handoff.md](references/handoff.md) · [references/observability.md](references/observability.md)
## Hard rules (enforced across everything the factory generates)
1. **Validate before running** — every ADW declares `REQUIRED_AGENTS` and calls `agents.validate()` first; a missing/misnamed agent fails before anything spawns.
2. **Typed outputs only** — every agent call pairs with a concrete `EnvelopeBase` subclass in `adw_modules/data_types.py`; parse failures re-prompt the same session (context intact), never restart.
**The output contract is a synced triad**: (a) the type in `data_types.py`, (b) the JSON example in the agent's `user.md` `## Report` section, (c) `output_type=` at every call site. These are ONE contract — change any one, update all three in the same edit (grep the type name to find every call site).
3. **Gates validate claims, not guesses** — `gate(envelope, run) -> list[str]` violations; failures return to the same session as corrections.
4. **Four-param rule** — any function with more than 4 parameters takes one concrete data type instead (`AgentCall`, `PhaseParams` are the pattern).
5. **One agent, one prompt, one purpose** — identity lives in `system.md`; task shape (user prompt + output type) lives at the call site.
6. **ADW scripts stay thin** — all low-level logic lives in `adw_modules/`.
7. **Every phase earns a description** — one sentence on what it does and why, never a restatement of its name. It is the only intent the trace, the console, and the UI ever show; `commit_plan: "Commit the plan"` is rejected at construction, blank is too.
8. **A known command is code, not an agent** — if you can write the invocation down (`bun test`, `ruff check`), it belongs in a `kind="code"` phase via `adw_modules/quality.py`. Agents are for the parts that need reading and deciding; failures come back to the builder as an envelope either way.
9. **`tools:` is a capability list, `writes:` is the boundary** — `bash` runs anything (including `git checkout`) and `write` reaches any path, so a tool list can never make "this agent changes nothing" true. `writes:` per agent and `protected_files` in defaults are enforced in `adw_modules/permissions.py` after every agent call: unauthorized changes are rolled back and the phase dies. The session runtime under `data_dir` is always writable — a read-only agent is read-only with respect to the REPO, never mute.
10. **Every ADW ends in `run.finish()`** — phases passing is not the same as the run being accepted. A test phase that ran a red suite succeeded at its job. Pass `accepted=` so the exit code, the session status, and the banner are decided together and cannot disagree.
## v1 scope
Pi coding agent only (`coding_agent: pi`), default model `gemini-3.6-flash` via openrouter, thinking `medium`. `claude_code` is schema-valid but stubbed until v2. The visualizer app ships in a later pass — observe via sqlite queries until then.
提问、分享技巧或报告无法使用。按“有帮助”数量排序。