workflow-ship-faster（工作流程发货更快）

Workflow: Ship Faster (Next.js 16.1.1)

The goal of this chain is: Ship an idea or small prototype to production-ready state in the shortest time, while maintaining iteration speed afterward.

When to use

Use this workflow when the user wants an end-to-end chain like:

• “ship this MVP”, “make it production-ready”, “launch/deploy this project”
• “take this prototype and turn it into a real app”
• “set up the foundation + UI system + deployment”

Core principles (non-negotiable)

• Pass paths only, not content: agents/sub-agents only pass ..._path.
• Files are first-class citizens: Every step must persist artifacts; failures can be retried; replayable.
• Confirmation points: Any "high-risk/high-effort/side-effect" action must write a plan first and wait for confirmation.
• Plans are checklists: progress is tracked in tasks.md via - [] → - [x] (not in chat).
• Progressive disclosure: Only open step files in this skill directory (foundation.md, deploy-vercel.md, etc.) when needed—avoid loading all details at once.
• Hooks doctor (required check; non-blocking): If running under Claude Code and you want the evolution loop active, run tool-hooks-doctor once at the start of the session; if hooks are missing, offer to install project-level hooks (continue either way).

Inputs / outputs (paths only)

Inputs (paths only):

• repo_root: project root (default ".")
• Optional: run_dir (active run directory) if the user already has one

Outputs (artifacts, written under run_dir/):

• Required: proposal.md, tasks.md, context.json
• Optional: design.md, evidence/, logs/

Canonical contract (backends, templates, read order, auto-archive):

• references/artifact-contract.md
• Reference index: references/README.md

Step modules (open only when you hit that step)

• Foundation: foundation.md
• Guardrails: guardrails.md
• Trace cleanup: cleanup-traces.md
• Docs baseline: docs-baseline.md
• Supabase integration: supabase-integration.md
• Stripe integration: stripe-integration.md
• Deploy (GitHub + Vercel): deploy-vercel.md
• AI-era SEO (Next.js): ai-seo-nextjs.md

Process (Default Route)

0) Initialize Run (Required)

1. Resolve/create run_dir/ (see the deterministic backend rules in references/artifact-contract.md).
2. Ensure core artifacts exist (create if missing; merge/append if already present):

- proposal.md, tasks.md, context.json

1. In tasks.md, ensure:

- a status: active|blocked|done field near the top - a short Next action section (1–3 items) - an Approvals section (empty is fine until needed)

1. Only create evidence/ and logs/ when you actually have large outputs to store.

0.2) Scope Confirmation (Required)

Problem this solves: a run summary that marks core steps as “skipped” without the user ever choosing to skip them is misleading and hard to audit.

Ask the user to select a scope before executing steps beyond build checks:

• A) full (default): Foundation → Design system → UI/UX → Guardrails → Docs → Feature iteration → Deploy
• B) deploy-only: Foundation (build + sanity) → Deploy only (no UI work, no guardrails/docs polish)
• C) design-only: Design system + UI/UX plan only (no code changes)
• D) feature-only: One feature iteration (plan + implementation), skip deploy

Rules:

• Persist the chosen scope to context.json (add/update a scope field).
• In tasks.md, do not mark a default step as “skipped” unless the user explicitly requested skipping it.

- Prefer: disabled (scope=...) with a short reason under a “Scope / Disabled steps” section

• Any destructive action (e.g., force push overwriting an existing repo) is a high-risk side effect:

- Write an executable approval item under tasks.md first - Wait for explicit user confirmation before executing

0.25) Kickoff Clarification (Brainstorm-lite) (Recommended; required if goal is vague)

Problem this solves: people jump straight into “implement a feature” and end up with a very basic MVP that’s hard to demo.

Run this step if any of these are true:

• proposal.md is missing clear acceptance criteria and non-goals
• The user request is “build something like X” / “make a prototype” without specifying the core loop
• The user explicitly wants a “demo-ready” prototype (animation, wow factor, shareable)

How:

• Call workflow-brainstorm using the same repo_root + this run_dir.
• Follow the one-question-at-a-time rule and converge on:

- the one core loop (1 sentence) - acceptance criteria (3–5 bullets) - non-goals (1–3 bullets) - constraints (timeline / risk preference) - a “demo moment” direction (see step 0.3)

• Persist the confirmed spec to:

- evidence/YYYY-MM-DD-kickoff-design.md

• Then update (merge, don’t overwrite) these inputs:

- proposal.md - context.json (ensure need_database/need_billing/need_deploy/need_seo are explicitly set)

0.3) Demo Moment First (Recommended for prototypes)

Principle: a prototype that feels real needs at least one “demo moment” (tastefully showy, not gimmicky).

Examples of “demo moment” (pick 1, keep it small):

• A hero section with a live interactive preview (fake data is fine) + smooth, purposeful motion
• A delightful micro-interaction: command palette, draggable cards, timeline scrubber, etc.
• A “before/after” transformation animation (input → output) that makes the core loop obvious

Workflow:

1. Write a small feature spec file (if it doesn’t exist):

- evidence/feature-00-demo-moment.md - Include mode: plan-only, feature_slug: demo-moment, and quality_bar: demo-ready

1. Call workflow-feature-shipper to generate the plan only (no implementation yet).
2. Defer actual UI implementation until after Step 2 (design-system.md exists), so the demo moment follows the chosen design system.

0.5) Dynamic Branch Decision (Required)

Dynamically adjust execution order based on context.json content:

• Database integration: Only execute when context.json has "need_database": true
• Payment integration: Only execute when context.json has "need_billing": true or "need_stripe": true
• Auth: Only execute when context.json has "need_auth": true or "auth": true
• Deployment: Only execute when context.json has "need_deploy": true or user explicitly requests deployment
• SEO: Only execute when context.json has "need_seo": true or project is already live

Before starting each optional step:

• Check context.json (and your chosen scope) to decide whether the step applies
• If skipped, write a 1-line reason into tasks.md (so resume/audit doesn’t require chat history)

1) Foundation: Next.js Foundation (Default Required)

Open and follow: foundation.md.

Artifact requirements (minimum):

• evidence/foundation.md (current state + risk assessment + chosen route)
• tasks.md: add a Foundation checklist section (tasks + verification commands)
• If upgrade/migration needed with significant changes: add an Approval item to tasks.md and wait for confirmation

Branch rules (important):

• If conclusion is keep-current-stack (e.g., Vite static site), don't force continuing "Next.js-specific steps".

- Only continue steps that still apply to current stack: style (2), guardrails (3, adapted for current stack), docs (4), deploy (8 optional), SEO (9 optional) - If DB/payment integration still needed: First ask user "want to migrate to Next.js?", or add integration step for that stack (not in this chain's default scope)

2) Style Specification (Default Required)

Call tool-design-style-selector: Scan project intent and persist design-system.md.

Confirmation point: After user confirms style, deploy design-system.md, and enter 2.5 UI/UX Implementation (default required; if user explicitly skips, log reason).

2.5) UI/UX Implementation (Default Required)

Goal: Make design-system.md actually reflected in the interface, not "wrote spec but UI unchanged".

Recommended approach: Treat "redo UI/UX per design-system.md" as an independent feature and hand to workflow-feature-shipper:

• Input: design-system.md (as sole design constraint) + current UI page list (from code scan)
• Output: evidence/features/<feature_slug>-plan.md (scope/acceptance criteria/non-goals/risks/rollback) + code changes

Default enrichment (when installed):

• If tool-ui-ux-pro-max is installed, use it to enrich the UI/UX plan with concrete palette/typography/UX guardrails, and use its pre-delivery checklist as acceptance criteria.
• Only skip tool-ui-ux-pro-max enrichment if the user explicitly asks to skip it (e.g., “don’t over-design / keep it simple”), and log the reason.

For large-scale refactoring: add an Approval item to tasks.md first and wait for user confirmation before implementing.

3) Code Standards (Lightweight Required)

Open and follow: guardrails.md.

Artifact: tasks.md (Guardrails checklist section)

3.5) Trace Cleanup (De-branding & Source Hygiene) (Required if project was copied)

Problem this solves: many starter projects imported from Google AI Studio / v0 / Lovable ship with vendor branding, broken remnants (like importmap), and misleading README instructions.

Open and follow: cleanup-traces.md.

Artifacts:

• evidence/trace-scan.md
• tasks.md (Trace cleanup checklist section)

4) Documentation Standards (Lightweight Required)

Open and follow: docs-baseline.md.

Artifact: tasks.md (Docs checklist section, plus project README update)

5) Core Feature Development (Iterative Loop)

Call workflow-feature-shipper:

• Each feature first produces evidence/features/<feature_slug>-plan.md (with acceptance criteria/non-goals)
• Default split into PR-able small steps
• After each batch (or before merge), recommend calling review-quality for a conclusive review + verdict
• review-quality is the single entry point and will auto-triage: if React/Next.js performance risk is detected, it will also run review-react-best-practices (CRITICAL rules first)

Auth note (optional): if a feature includes login/session/permissions and tool-better-auth is installed, call it before implementation to lock down session strategy, redirects, cookies/CSRF, and middleware boundaries. Persist:

• evidence/auth-plan.md
• evidence/auth-summary.md

6) Database Integration (Optional)

Execution condition: context.json has "need_database": true or "database": "supabase"

Open and follow: supabase-integration.md.

7) Payment Integration (Optional)

Execution condition: context.json has "need_billing": true or "need_stripe": true or "payment": true

Open and follow: stripe-integration.md.

8) GitHub + Vercel Deployment (Optional but Recommended)

Execution condition: context.json has "need_deploy": true or "deploy_target" field exists and non-empty

Open and follow: deploy-vercel.md.

9) AI-Era SEO (Optional but Recommended)

Execution condition: context.json has "need_seo": true or "seo": true, or step 8 completed successfully (already deployed)

Open and follow: ai-seo-nextjs.md.

Final Delivery

Write to: final.md (in the run root), at minimum include:

• Current project status (can run locally, can build, is deployed)
• Completed steps and artifact paths
• Next steps (prioritized by value)

And wrap up:

• Update tasks.md status to done, and fill the Delivery summary
• Run auto-archive after each batch (and at the end):

- python3 ~/.claude/skills/workflow-ship-faster/scripts/auto_archive.py --run-dir "<run_dir>"

• If you want a clean workspace (without auto-archive): move active/<run_id>/ → archive/YYYY-MM-DD-<run_id>/
• Do a skill-evolution Evolution checkpoint (3 questions); if user chooses "want to optimize", run skill-improver based on this run_dir to produce minimal patch suggestions