菜单
Defines what is being built before any code is written. Use when a Kanban card enters the backlog without acceptance criteria, when starting a new feature or significant change, or when the requirements are ambiguous enough that two engineers could implement the task differently.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/howmon/fauna --skill spec-driven-development命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Create editorial architecture, flowchart, sequence, ER, state machine, timeline, swimlane, quadrant, radar, loop, nested, tree, org chart, layer stack, Venn, pyramid, bar, line, Gantt, scatter, high-level, process, medallion, data flow, IT current-state, DP integration, and DP security matrix diagrams — plus fauna-native circuit schematics and PCB board views — as standalone self-contained HTML/SVG artifacts. Inherits fauna's active theme preset. Redraw draw.io / Mermaid sources; add brand tokens from a website.
Write or rewrite any text — technical docs, articles, emails, conversations, README files, runbooks, error messages, release notes, incident reports, or chat messages — with the rules of ASD-STE100 Simplified Technical English. Produces clear, unambiguous, slop-free prose. Use whenever the user asks to write, draft, rewrite, or improve ANY text. Also triggers on: "STE", "Simplified Technical English", "ASD-STE100", "de-slop", "simple english", "make this readable", "write clearly", "write for non-native readers". Enforces 53 rules: 20/25-word sentence limits, one word one meaning, simple tenses, active voice, no hedging, condition before command.
Performs multi-axis code review before merge. Use when a Kanban card enters the review column, when evaluating code another agent produced, or before merging any change. Reviews across correctness, readability, architecture, security, and performance.
| name | spec-driven-development |
| description | Defines what is being built before any code is written. Use when a Kanban card enters the backlog without acceptance criteria, when starting a new feature or significant change, or when the requirements are ambiguous enough that two engineers could implement the task differently. |
Before writing any non-trivial code, write a short spec covering the objective, the user-visible behaviour, the boundaries (what is and isn't in scope), and the acceptance criteria. A spec is cheap to revise. Code built on the wrong spec is expensive to revise.
When NOT to use: Bug fixes with a clear reproduction. Typo fixes. Dependency bumps.
Write a spec with these sections (keep it to one screen — this is a contract, not a design doc):
One sentence: what user outcome does this produce? Not "add a button" — "let the user create a folder inside a project source so they can organise files without leaving Fauna".
Numbered list of what the user will see, click, or get back. Concrete enough that you could write a UI test from it.
Bulleted list of testable conditions. Each one becomes a verification item later. Format as Given/When/Then or "MUST/SHOULD/MUST NOT".
The single most-skipped section. Name 2-3 things you could reasonably do here but won't, with one-line reasons. Prevents scope creep mid-task.
If anything is ambiguous, list it here with your best-guess answer and a "correct me now or I'll proceed with this" note.
| Rationalization | Reality |
|---|---|
| "The task title is enough" | Then writing a one-screen spec from it takes 60 seconds. Do it. |
| "I'll figure out the requirements as I code" | You will not. You will figure out your own preferences and call them requirements. |
| "The user will tell me if I get it wrong" | After 30 minutes of work. Spec first. |
| "The spec will lock me in" | The spec is revisable. The code built on the wrong spec is not. |
Before leaving this skill: