sdd-refine
SDD step 2 (optional, repeatable). Read an existing feature.md and refine it by updating or enhancing requirements based on user input. Use when requirements have changed, new edge cases are discovered, or the spec needs clarification before planning or re-planning. Run before /sdd-plan if plan.md already exists.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 18
- Forks
- 6
01 前言 Front Matter
- name
- sdd-refine
- description
- SDD step 2 (optional, repeatable). Read an existing feature.md and refine it by updating or enhancing requirements based on user input. Use when requirements have changed, new edge cases are discovered, or the spec needs clarification before planning or re-planning. Run before /sdd-plan if plan.md already exists.
- argument-hint
- <what to change or enhance (optional)>
02 资源文件 1
- 参考文档skill.json423 B
03 SKILL.md 正文
SDD: Refine Feature Spec
You are a senior software architect refining an existing feature specification.
Inputs
| Input | Required | Description | Example |
|---|---|---|---|
refinement_request |
Optional | What to change or add to the spec | "Add rate limiting: max 5 login attempts per minute" |
Steps
Step 0: Validate Inputs (ALWAYS DO THIS FIRST)
Check the conversation for refinement_request and for the spec directory (docs/specs/<yyyymmddHHMM>-<feature-name>/, the one holding feature.md). If several are open, ask which one.
- If
feature.mddoes not exist → stop and tell the user to run/sdd-featurefirst. - If
refinement_requestis present → proceed to Step 1. - If
refinement_requestis missing → ask:"What would you like to change or add to the spec? Is this a scope change, a clarification, or new edge cases?" Do NOT proceed until the user provides it.
Pre-conditions
Verify feature.md exists in the spec directory.
If it does not exist, tell the user to run /sdd-feature first.
1. Read Current State
Read these files before doing anything:
feature.md: the existing spec to be refineddocs/project.md: project context and constraintsplan.md: if it exists, note which parts of the plan may be invalidated by changes
2. Understand the Refinement Request
The refinement input is: refinement_request (collected in Step 0).
Analyse refinement_request against the current feature.md and identify:
- What sections are affected
- Whether the change expands scope, reduces scope, or clarifies existing scope
- Any knock-on effects (e.g., changing a requirement may invalidate other ACs)
3. Clarify One Question at a Time
Open [NEEDS CLARIFICATION: ...] markers in feature.md come first, then any ambiguity in the
refinement request itself. Work through them one by one:
- Ask one question. Offer 2 to 4 concrete options and mark one as recommended, with the reason in one line (impact, scope, risk).
- Wait for the answer.
- Apply it immediately to the section it affects, and delete the marker it resolves.
- Only then ask the next question. Stop after 5 questions in one session; anything left stays as a marker.
4. Show a Diff Summary Before Editing
The answers of step 3 are already in the file. For every other change the refinement asks for, present a brief plan before modifying the file:
## Proposed Changes to feature.md
### Additions
- FR-04: <new requirement>
- AC-05: <new acceptance criterion>
### Modifications
- FR-02: Updated to clarify that X also applies to Y
- AC-02: Strengthened: must complete within 200ms, not 500ms
### Removals
- FR-03: Removed: out of scope per user confirmation
### No Change
- All other sections remain as-is
Ask the user to confirm before applying.
5. Apply the Refinements
Update feature.md in place. Preserve:
- Existing section structure and numbering where possible
- The checkbox state of every criterion exactly as found; never tick or untick one
- The
Spec-Formatheader, the Glossary, and EARS form for every criterion (new criteria follow the patterns in/sdd-feature, with a Glossary name as subject) - Open markers you did not resolve; add a new one only if the refinement raises a question you cannot answer, and never exceed 3 in total
Increment requirement IDs sequentially (do not reuse deleted IDs).
6. Impact Assessment
After updating feature.md, check if plan.md exists.
If it does, analyse the impact:
## Impact on plan.md
plan.md exists and may be partially invalidated. Here is what needs revisiting:
- Step 2 (Domain Layer): FR-04 adds a new value object not currently planned
- Step 6 (Tests): 2 new ACs require additional test cases
- Step 3 is unaffected
Recommendation: Run /sdd-plan again to regenerate the plan before implementing.
If plan.md does not exist, simply confirm the spec is updated and prompt the user to run /sdd-plan.
7. Changelog Entry
Append a refinement record at the bottom of feature.md:
---
## Revision History
| Date | Change Summary |
|------|----------------|
| <date> | Initial spec |
| <date> | <One-line summary of this refinement> |
If a revision history table already exists, append a new row; do not recreate the table. The row names the questions answered in this session. This table is the only clarification log.
8. Lint
Run python3 ~/.claude/scripts/spec_lint.py <spec-directory> and fix every finding.
04 同仓库更多 Skills
yeet
Use only when the user explicitly asks to stage, commit, push, and open a GitHub pull request in one flow using the GitHub CLI (`gh`).
wrangler
Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
workspace-skill-discovery
Discover and include ALL skills across the workspace — not just global ones. Use at the start of every session or when the user asks about available skills. Ensures project-level skills (under .claude/skills/ or skills/ in any repo) are never missed.
workers-best-practices
Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
winui-app
Create, build and design WinUI 3 desktop apps with C# and the Windows App SDK following Microsoft's official guidance and CommunityToolkit. Covers XAML, navigation, theming, accessibility and deployment.
windows-brain-script-portability
Scripts in ~/.claude/scripts/ that run on a Windows brain: use the python3.cmd shim and not bare python3, and reconfigure stdout to UTF-8 if you print symbols. Use it when writing a brain script or triaging a Windows-only crash.