devops-infra-scaffold（DevOps 基础架构）

devops-infra-scaffold

Generate CI/CD workflows, Dockerfiles, and deployment configs from tech.md — you can't ship what you can't deploy.

Context Files

• $JAAN_CONTEXT_DIR/tech.md - Tech stack context (CRITICAL -- determines framework, services, deployment target)

- Uses sections: #current-stack, #frameworks, #constraints, #versioning

• $JAAN_CONTEXT_DIR/config.md - Project configuration
• $JAAN_TEMPLATES_DIR/jaan-to-devops-infra-scaffold.template.md - Output template
• $JAAN_LEARN_DIR/jaan-to-devops-infra-scaffold.learn.md - Past lessons (loaded in Pre-Execution)
• ${CLAUDE_PLUGIN_ROOT}/docs/extending/language-protocol.md - Language resolution protocol
• ${CLAUDE_PLUGIN_ROOT}/docs/research/74-dev-cicd-infra-scaffold-generation.md - Research reference

Input

Upstream Artifacts: $ARGUMENTS

Accepts file paths or descriptions:

• tech.md -- Path to tech stack definition (from $JAAN_CONTEXT_DIR/tech.md)
• backend-scaffold output -- Path to backend scaffold output (from /jaan-to:backend-scaffold)
• frontend-scaffold output -- Path to frontend scaffold output (from /jaan-to:frontend-scaffold)
• detect-dev output -- Path to detect-dev output (optional, from /jaan-to:detect-dev)
• Empty -- Interactive wizard prompting for tech stack, CI/CD platform, and deployment target

Pre-Execution Protocol

MANDATORY — Read and execute ALL steps in: ${CLAUDE_PLUGIN_ROOT}/docs/extending/pre-execution-protocol.md Skill name: devops-infra-scaffold Execute: Step 0 (Init Guard) → A (Load Lessons) → B (Resolve Template) → C (Offer Template Seeding)

Also read context files if available:

• $JAAN_CONTEXT_DIR/tech.md -- Know the tech stack for framework-specific infrastructure generation
• $JAAN_CONTEXT_DIR/config.md -- Project configuration

Language Settings

Read and apply language protocol: ${CLAUDE_PLUGIN_ROOT}/docs/extending/language-protocol.md Override field for this skill: language_devops-infra-scaffold

Language exception: Generated code output (Dockerfiles, YAML workflows, shell scripts,.env files, deployment configs) is NOT affected by this setting and remains in English/code.

PHASE 1: Analysis (Read-Only)

Thinking Mode

ultrathink

Use extended reasoning for:

• Analyzing tech stack to determine optimal CI/CD pipeline stages
• Mapping framework + database + deployment target to infrastructure patterns
• Planning multi-stage Docker build strategy from detected stack
• Identifying environment variable hierarchy and secret management needs
• Evaluating existing CI/CD setup from detect-dev output (if available)

Step 1: Parse Tech Stack & Upstream Artifacts

Read and parse all available inputs:

1. tech.md -- Extract from #current-stack:

- Languages and frameworks (Node.js/Next.js, PHP/Laravel, Go, etc.) - Databases (PostgreSQL, MySQL, Redis, etc.) - Package manager (pnpm, npm, yarn, composer, go mod) - Monorepo tool (Turborepo, Nx, none)

1. backend-scaffold output -- Extract:

- Entry points and build commands - Dependencies and ORM - Port configuration - Environment variables referenced

1. frontend-scaffold output -- Extract:

- Framework and build config (Next.js standalone, Vite, etc.) - Static vs SSR rendering - Build output directory - Environment variables (public + server)

1. detect-dev output (optional) -- Extract:

- Existing CI/CD workflows - Current Dockerfiles - Existing deployment configs - Identified gaps and recommendations

Present input summary:

INPUT SUMMARY
-------------
Tech Stack:       {framework} + {database} + {cache}
Package Manager:  {package_manager}
Monorepo:         {monorepo_tool or "no"}
Backend Entry:    {entry_point}
Frontend Build:   {build_tool / output_mode}
Existing CI/CD:   {found / none}
Sources Found:    {list}
Sources Missing:  {list with fallback suggestions}

Step 2: Clarify Infrastructure Decisions

AskUserQuestion for items not derivable from inputs:

• CI/CD Platform: GitHub Actions (default) / GitLab CI / other
• Deployment Target: Vercel (frontend) + Railway (backend) / Fly.io / AWS ECS / Docker Compose only
• Container Registry: GitHub Container Registry (ghcr.io, default) / Docker Hub / AWS ECR / none
• Environment Strategy: How many environments? (dev / staging / production)
• Database Migrations: Which tool? (Prisma / Drizzle / Knex / golang-migrate / Laravel migrations)
• Security Scanning: Include Trivy container scanning? (recommended: yes)

Step 3: Plan Infrastructure Structure

Based on tech stack + decisions, plan the complete infrastructure scaffold:

INFRASTRUCTURE PLAN
===================

CI/CD PLATFORM: {platform}
DEPLOYMENT:     {target}
REGISTRY:       {registry}
ENVIRONMENTS:   {list}

OUTPUT STRUCTURE
----------------
{id}-{slug}/
+-- {id}-{slug}.md                          # Infrastructure guide
+-- ci/
|   +-- ci.yml                              # CI workflow (lint, type-check, test, build)
|   +-- cd.yml                              # CD workflow (deploy to environments)
|   +-- health-check.yml                    # Health monitoring (cron, 15-min interval)
|   +-- secret-rotation-reminder.yml        # Secret rotation reminder (quarterly)
+-- docker/
|   +-- Dockerfile.backend                  # Multi-stage backend build
|   +-- Dockerfile.frontend                 # Multi-stage frontend build
|   +-- docker-compose.yml                  # Full-stack dev environment
|   +-- docker-compose.prod.yml             # Production overrides
|   +-- .dockerignore                       # Build context exclusions
+-- config/
|   +-- .env.example                        # All variables with safe defaults
|   +-- .env.test                           # Test environment variables
|   +-- .env.production.example             # Production template (no secrets)
|   +-- next.config.standalone.ts           # Next.js standalone config (if Next.js detected)
+-- deploy/
|   +-- {platform}.yml                      # Platform-specific config
|   +-- migration.sh                        # Database migration script
+-- {id}-{slug}-readme.md                   # Setup + deployment instructions

PIPELINE STAGES
---------------
CI: {stage_list}
CD: {stage_list}
Docker Stages: {stage_list}
Services: {service_list}

Report any conflicts or missing information.

HARD STOP -- Review Infrastructure Plan

Use AskUserQuestion:

• Question: "Proceed with generating the infrastructure scaffold?"
• Header: "Generate"
• Options:

- "Yes" -- Generate all infrastructure files - "No" -- Cancel - "Edit" -- Let me revise the deployment target or CI/CD strategy first

Do NOT proceed to Phase 2 without explicit approval.

PHASE 2: Generation (Write Phase)

Phase 2 Output -- Folder with subfolders

All files in $JAAN_OUTPUTS_DIR/devops/infra-scaffold/{id}-{slug}/:

{id}-{slug}/
+-- {id}-{slug}.md                          # Main doc (infrastructure guide)
+-- ci/
|   +-- ci.yml                              # GitHub Actions CI workflow
|   +-- cd.yml                              # GitHub Actions CD workflow
|   +-- health-check.yml                    # Health monitoring (cron, 15-min interval)
|   +-- secret-rotation-reminder.yml        # Secret rotation reminder (quarterly)
+-- docker/
|   +-- Dockerfile.backend                  # Multi-stage backend Dockerfile
|   +-- Dockerfile.frontend                 # Multi-stage frontend Dockerfile
|   +-- docker-compose.yml                  # Development docker-compose
|   +-- docker-compose.prod.yml             # Production overrides (optional)
|   +-- .dockerignore                       # Build context exclusions
+-- config/
|   +-- .env.example                        # All env vars with safe defaults
|   +-- .env.test                           # Test environment config
|   +-- .env.production.example             # Production template (no secrets)
|   +-- next.config.standalone.ts           # Next.js standalone config (if Next.js detected)
+-- deploy/
|   +-- {platform}.yml                      # Deployment platform config
|   +-- migration.sh                        # Database migration script
+-- {id}-{slug}-readme.md                   # Setup + deployment instructions

Step 5: Generate CI Workflow (ci.yml)

Generate GitHub Actions CI workflow with these stages:

1. Detect Changes -- Use dorny/paths-filter@v3 for monorepo path filtering
2. Lint -- ESLint/Biome (Node.js), PHP-CS-Fixer (PHP), golangci-lint (Go)
3. Type Check -- tsc --noEmit (TypeScript), PHPStan (PHP), go vet (Go)
4. Test -- With service containers (PostgreSQL, Redis) and healthchecks
5. Build -- Framework-specific build with caching
6. Security Scan -- Trivy filesystem scan + pnpm audit / npm audit

Caching strategy:

• Package manager cache via actions/setup-node@v4 (or equivalent)
• Build output cache via actions/cache@v4 (Next.js .next/cache, TypeScript tsbuildinfo)
• Docker layer cache via docker/build-push-action@v5 with cache-from: type=gha

Key patterns from research:

• fail-fast: false for matrix builds
• Reusable workflow structure for DRY CI
• retention-days: 1 for ephemeral build artifacts
• Pin actions by SHA for supply chain security

pnpm packageManager Conflict Prevention

When tech.md #current-stack indicates pnpm as package manager, check the project's package.json for a packageManager field:

• If packageManager field exists (e.g., "packageManager": "pnpm@9.x.x"): Use pnpm/action-setup@v4 without the version parameter — the action reads version from package.json automatically. Explicit version causes ERR_PNPM_BAD_PM_VERSION.
• If packageManager field is absent: Use pnpm/action-setup@v4 with explicit version parameter.

Applies to both ci.yml and cd.yml.

Step 6: Generate CD Workflow (cd.yml)

Generate deployment workflow triggered on:

• Push to main -- Deploy to production
• Push to develop -- Deploy to staging (if configured)
• PR -- Deploy preview (if platform supports it)

Stages:

1. Build Docker Images -- Multi-stage build, push to registry
2. Run Migrations -- Separate job, before deployment
3. Deploy -- Platform-specific (Vercel CLI / Railway CLI / Fly deploy / ECS update)
4. Smoke Test -- Health check on deployed URL
5. Notify -- Success/failure notification (optional)

Environment protection:

• Use GitHub Environments with required reviewers for production
• Use OIDC federation for cloud credentials (no stored secrets)
• Separate secrets per environment

Step 6a: Generate Optional Quality Pipeline Stages

Generate these stages ONLY if relevant config/spec is detected in the project. Each stage checks tool availability using npx --no-install (npm tools) or direct --version (binary tools). Never use bare npx (auto-installs). Never use which.

6a.1 Spectral Lint Stage (if OpenAPI spec detected)

Glob for *.yaml/*.json matching OpenAPI patterns (openapi:, swagger:). If found:

- name: API Spec Lint (Spectral)
  run: npx --no-install @stoplight/spectral-cli lint api.yaml --ruleset .spectral.yaml

• Trigger: on every commit (CI workflow)
• Ruleset: built-in + OWASP if .spectral.yaml exists

6a.2 Breaking Changes Detection Stage (if baseline spec exists)

If project has a baseline OpenAPI spec (e.g., api.yaml tracked in git):

- name: API Breaking Changes (oasdiff)
  uses: oasdiff/oasdiff-action@{pinned-sha}
  with:
    base: api.yaml
    revision: api.yaml
    fail-on: ERR

• oasdiff is a Go binary, NOT npm. CI uses oasdiff/oasdiff-action pinned to immutable commit SHA (never @latest).
• Local: oasdiff breaking --fail-on ERR base.yaml head.yaml
• Trigger: on PRs only

6a.3 Mutation Testing Stage (if mutation tool config detected)

Glob for mutation configs: stryker.config.*, infection.json5, .mutmut-cache. If found:

# PR: incremental (changed files only)
- name: Mutation Testing (Incremental)
  if: github.event_name == 'pull_request'
  run: npx stryker run --incremental

# Nightly: full run
- name: Mutation Testing (Full)
  if: github.event_name == 'schedule'
  run: npx stryker run

• Adapt command per stack (StrykerJS / Infection / go-mutesting / mutmut)
• PR runs: incremental, changed files only
• Nightly runs: full project scope

6a.4 API Fuzz Testing Stage (if OpenAPI spec + running API)

If OpenAPI spec detected AND deployment target configured:

- name: API Fuzz Testing (Schemathesis)
  run: schemathesis run --url ${{ vars.API_URL }} api.yaml --stateful=links

• Trigger: nightly/on-demand only (resource-intensive)
• schemathesis is a Python pip package, NOT npm

6a.5 Preflight Availability Gate

Each stage above MUST include a preflight check before execution:

• npx --no-install @stoplight/spectral-cli --version for Spectral
• oasdiff --version for oasdiff (binary, not npm)
• npx --no-install stryker --version for StrykerJS (adapt per stack)
• schemathesis --version for Schemathesis

If tool unavailable: skip stage with explicit comment in generated workflow, never silently omit.

Step 7: Generate Health Check Workflow (health-check.yml) — GitHub Actions Only

Platform conditional: Skip this step when CI platform is GitLab CI.

Generate a scheduled health monitoring workflow:

• Triggers: Cron every 15 min (*/15 * * * *) + workflow_dispatch for manual testing
• Endpoint checks: curl with 30s timeout against configurable endpoints
• Endpoint URLs: Stored as repository variables (vars.API_URL, vars.WEB_URL) — not secrets
• On failure: actions/github-script (pinned by SHA) creates issue with incident label
• Deduplication: Search open issues with incident label before creating — skip if duplicate exists
• On recovery: Auto-close open incident issue with resolution comment and timestamp
• Stack-agnostic: Works for any tech stack with HTTP endpoints (Node.js, PHP, Go, etc.)

Step 8: Generate Secret Rotation Reminder (secret-rotation-reminder.yml) — GitHub Actions Only

Platform conditional: Skip this step when CI platform is GitLab CI.

Generate a quarterly secret rotation reminder workflow:

• Trigger: Quarterly cron 0 9 1 1,4,7,10 * (Jan/Apr/Jul/Oct, 9 AM UTC)
• Action: actions/github-script (pinned by SHA) creates issue with security + maintenance labels
• Checklist body: Generated from project's .env.example / .env.production.example:

- Rotate (credentials): Variables containing TOKEN, SECRET, KEY, PASSWORD, or URL-with-credentials - Static (no rotation needed): Variables containing ID, NAME, REGION, ENV, PORT

• Stack-agnostic: Classification heuristic works for any project's environment variables

Step 9: Generate Dockerfiles

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Dockerfile Generation Patterns" for per-stack Dockerfile patterns (backend, frontend,.dockerignore).

Next.js Standalone Output Consistency

When tech.md #current-stack indicates Next.js and generating Dockerfile.frontend with standalone COPY stage:

1. Check if project's next.config.ts (or .js/.mjs) already has output: 'standalone'
2. If missing: generate config/next.config.standalone.ts reference snippet showing the required config
3. Add README step: "Verify output: 'standalone' in next.config.ts — required for Docker multi-stage build"

Step 10: Generate docker-compose.yml

Full-stack development environment with:

Services (based on tech.md):

• Backend app (build from Dockerfile.backend, target: development)
• Frontend app (build from Dockerfile.frontend, target: development)
• PostgreSQL (if detected) -- with healthcheck (pg_isready)
• MySQL (if detected) -- with healthcheck (mysqladmin ping)
• Redis (if detected) -- with healthcheck (redis-cli ping)
• Additional services from tech.md

Patterns from research:

• condition: service_healthy for all depends_on
• Named volumes for database persistence
• Bind mounts for hot-reload (source code only)
• Anonymous volumes to protect node_modules
• Network isolation (frontend / backend networks)
• Profiles for selective startup (--profile backend, --profile full)
• Environment variables from .env file

docker-compose.prod.yml (optional overlay):

• Production-optimized settings
• No bind mounts
• Resource limits
• Restart policies

Step 11: Generate Environment Config Files

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Environment Config File Templates" for .env.example, .env.test, and .env.production.example templates.

Generate three env config files based on the templates in the reference, customized to the detected tech stack and services from tech.md.

Step 12: Generate Deployment Config

Based on selected deployment target:

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Deployment Platform Configurations" for Vercel, Railway, Fly.io, AWS ECS configs and migration.sh.

Step 13: Quality Check

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Quality Checklist" for the full validation checklist.

Validate generated output against the quality checklist and output structure checklist in the reference. If any check fails, fix before preview.

Step 14: Preview & Approval

Present generated output summary showing:

• File count and structure
• CI stages and estimated pipeline time
• Docker image target sizes
• Environment variable count
• Deployment target summary

Use AskUserQuestion:

• Question: "Write infrastructure scaffold files to output?"
• Header: "Write Files"
• Options:

- "Yes" -- Write the files - "No" -- Cancel - "Refine" -- Make adjustments first

Step 15: Generate ID and Folder Structure

source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/id-generator.sh"
SUBDOMAIN_DIR="$JAAN_OUTPUTS_DIR/devops/infra-scaffold"
mkdir -p "$SUBDOMAIN_DIR"
NEXT_ID=$(generate_next_id "$SUBDOMAIN_DIR")
slug="{project-name-slug}"
OUTPUT_FOLDER="${SUBDOMAIN_DIR}/${NEXT_ID}-${slug}"

Preview output configuration:

Output Configuration - ID: {NEXT_ID} - Folder: $JAAN_OUTPUTS_DIR/devops/infra-scaffold/{NEXT_ID}-{slug}/ - Main file: {NEXT_ID}-{slug}.md

Step 16: Write Output

1. Create output folder and subfolders:

mkdir -p "$OUTPUT_FOLDER"
mkdir -p "$OUTPUT_FOLDER/ci"
mkdir -p "$OUTPUT_FOLDER/docker"
mkdir -p "$OUTPUT_FOLDER/config"
mkdir -p "$OUTPUT_FOLDER/deploy"

1. Write all files to respective subfolders:

- {id}-{slug}.md -- Main infrastructure guide (from template) - ci/ci.yml -- CI workflow - ci/cd.yml -- CD workflow - ci/health-check.yml -- Health monitoring workflow (GitHub Actions only) - ci/secret-rotation-reminder.yml -- Secret rotation reminder (GitHub Actions only) - docker/Dockerfile.backend -- Backend Dockerfile - docker/Dockerfile.frontend -- Frontend Dockerfile - docker/docker-compose.yml -- Dev docker-compose - docker/docker-compose.prod.yml -- Production overrides (if applicable) - docker/.dockerignore -- Build context exclusions - config/.env.example -- All env vars - config/.env.test -- Test env vars - config/.env.production.example -- Production template - config/next.config.standalone.ts -- Next.js standalone config (if Next.js detected) - deploy/{platform}.yml -- Deployment config - deploy/migration.sh -- Migration script - {id}-{slug}-readme.md -- Setup + deployment instructions

1. Update subdomain index:

source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/index-updater.sh"
add_to_index \
  "$SUBDOMAIN_DIR/README.md" \
  "$NEXT_ID" \
  "${NEXT_ID}-${slug}" \
  "{Project Title} Infrastructure" \
  "{Executive summary -- 1-2 sentences}"

1. Confirm completion:

Scaffold written to: $JAAN_OUTPUTS_DIR/devops/infra-scaffold/{NEXT_ID}-{slug}/ Index updated: $JAAN_OUTPUTS_DIR/devops/infra-scaffold/README.md

Step 17: Suggest Next Actions

Infrastructure scaffold generated successfully! Next Steps: - Copy CI/CD workflows to .github/workflows/ - Copy Dockerfiles and docker-compose to project root - Copy .env.example to project root and create .env from it - Copy deployment config to project root - Run docker compose up to verify local development environment - Push a branch to test CI workflow - Configure endpoint repository variables (API_URL, WEB_URL) for health monitoring workflow - Verify output: 'standalone' in next.config.ts (if using Next.js Docker build) - Run /jaan-to:sec-audit-remediate to audit security of generated configs - Run /jaan-to:learn-add devops-infra-scaffold "{feedback}" to capture lessons

Step 18: Capture Feedback

Use AskUserQuestion:

• Question: "How did the infrastructure scaffold turn out?"
• Header: "Feedback"
• Options:

- "Perfect!" -- Done - "Needs fixes" -- What should I improve? - "Learn from this" -- Capture a lesson for future runs

If "Learn from this": Run /jaan-to:learn-add devops-infra-scaffold "{feedback}"

Multi-Stack Support (Research-Informed)

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Multi-Stack Infrastructure Patterns" for detection table and per-stack key patterns (Node.js, PHP/Laravel, Go).

Security Best Practices (Applied to All Generated Files)

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Security Best Practices" for full security checklist applied to all generated files.

DAG Position

tech.md + backend-scaffold + frontend-scaffold + detect-dev (optional)
  |
  v
devops-infra-scaffold

Skill Alignment

• Two-phase workflow with HARD STOP for human approval
• Multi-platform support (GitHub Actions, GitLab CI, etc.)
• Template-driven output structure
• Output to standardized $JAAN_OUTPUTS_DIR path

Definition of Done

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/devops-infra-scaffold-reference.md section "Definition of Done" for the full checklist.