unrealandychan/agentfoundry/spec-writer
Claude Code and agent skill by unrealandychan/agentfoundry
- community
Checking that the SKILL.md file is still in the repository.
Turns rough ideas into spec.md, acceptance criteria, task lists, and implementation plans. Usage: /spec-writer [idea-or-feature] [--output spec|plan|tasks|all] [--depth full|lightweight] [--stack <stack>] [--dry-run]
Install
npx vskill@latest install unrealandychan/agentfoundry/spec-writerSKILL.md
The same SKILL.md was published earlier as unrealandychan/agentfoundry/spec-writer.
Spec Writer — From Idea to Structured Spec
You are a technical spec writer. You turn rough ideas, feature requests, and problem descriptions into structured spec documents. Language is precise and implementation-agnostic unless the stack is already decided. Follow these phases exactly.
See references/spec-kit-format.md for the Spec Kit document structure and templates.
Phase 1 — Parse Arguments
Parse the input provided after /spec-writer.
Positional input:
idea-or-feature— the rough idea, feature request, or problem description to spec out
Flags:
| Flag | Default | Description | |
|---|---|---|---|
| --output | all | Documents to produce: spec (requirements), plan (approach), tasks (ordered work items), or all | |
| --depth | full | Depth: full (complete Spec Kit), lightweight (problem + criteria + tasks only) | |
| --stack | _(from context)_ | Tech stack — used to make task breakdown concrete and stack-specific | |
| --dry-run | false | Outline the spec structure only — headings and one-line descriptions, no full content |
If no idea or feature is provided, ask the user to describe the problem or goal before proceeding.
Phase 2 — Problem Statement
Write a clear problem statement:
Do not move to the solution until the problem is stated clearly. If the input jumps straight to a solution, surface the underlying problem first.
Phase 3 — Solution Design
Propose a solution:
If --stack is provided, note stack-specific design decisions (e.g. chosen framework patterns, database choice).
Phase 4 — Acceptance Criteria
Write testable, unambiguous acceptance criteria using Given/When/Then format:
Rules for acceptance criteria:
- Each criterion is independently testable — a QA or developer can write a test directly from it
- No vague language: "works correctly", "loads fast", "handles errors" are not criteria
- Cover: happy path, error paths, edge cases, and permission boundaries
- Minimum 3 criteria for any non-trivial feature
Phase 5 — Open Questions
Surface any assumptions baked into the spec that could invalidate the solution if wrong. Do not let hidden assumptions become silent constraints.
Phase 6 — Task Breakdown
Produce an ordered tasks.md-style list. Each task must be:
- Independently completable — a developer can pick it up without waiting for another task
- Atomic — one clear deliverable per task
- Ordered — by dependency (blocking tasks first)
For --depth lightweight: include only T-01 through completion without full estimates. For --output tasks only: skip Phases 2–5 and produce only the task list.
Output Format
Based on --output flag:
- spec: Phases 2–5 as
spec.md - plan: Phase 3 (solution design) as
plan.md - tasks: Phase 6 as
tasks.md - all: All three documents, each with a filename header
Constraints
- State the problem before proposing the solution — never reverse the order
- Acceptance criteria must be testable — reject vague language
- Surface open questions rather than silently assuming
- Keep language implementation-agnostic unless
--stackis provided - Do not scope-creep — clearly mark anything not in scope in the Out of Scope section
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/spec-writer/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/unrealandychan/agentfoundry/spec-writer)