Spec-driven development workflow. Research your codebase, create implementation plans, implement from the plan, and review what was built. Includes 6 specialist agents and 5 workflow commands.
Review recent code changes for quality, security, and correctness. Run before committing.
Create an implementation plan interactively. Run after research, before implementing.
Implement an approved plan phase by phase. Run after create-plan.
Research the codebase and produce research.md. Run this before planning anything.
Transform a "do X" prompt into a research-focused prompt. Run this when you catch yourself about to build before understanding.
Use this agent when code has been written and needs review before committing. This agent should be called proactively after any significant code changes, new features, bug fixes, or refactoring work.
Analyzes codebase implementation details. Call the codebase-analyzer agent when you need to find detailed information about specific components. As always, the more detailed your request prompt, the better! :)
Locates files, directories, and components relevant to a feature or task. Call `codebase-locator` with human language prompt describing what you're looking for. Basically a "Super Grep/Glob/LS tool" — Use it if you find yourself desiring to use one of these tools more than once.
codebase-pattern-finder is a useful subagent_type for finding similar implementations, usage examples, or existing patterns that can be modeled after. It will give you concrete code examples based on what you're looking for! It's sorta like codebase-locator, but it will not only tell you the location of files, it will also give you code details!
Use this agent when the user wants to run tests, execute the test suite, verify code changes with tests, check test coverage, or validate functionality through testing. This includes running all tests, specific test files, or individual test methods. Also use when the user wants to verify that code changes haven't broken existing functionality.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Plugin marketplace for the AI Engineering Workshop, Part 2: Spec-Driven Workflow.
/plugin marketplace add github:Visma-Tech-LT/ai-workshop-marketplace
/plugin install sdw@ai-workshop
That's it. You now have 6 agents and 5 workflow commands available.
/sdw:research-codebase → /sdw:create-plan → /sdw:implement-plan → /sdw:code-review
Run each command, review its output, then proceed to the next. This is intentional — explained below.
1. Research — understand what exists before building anything
/sdw:research-codebase
Tell Claude what you want to understand. It explores the codebase and outputs research.md. Review it, then /clear.
2. Plan — create a detailed, phased implementation plan
/sdw:create-plan based on @.claude/ai/research-YYYY-MM-DD-topic.md
Interactive process. Claude asks questions, proposes phases, writes plan.md. Review it, then /clear.
3. (Optional) Skeptic Gate
Review this plan. Challenge it.
Claude adversarially reviews the plan. Fix weak spots before implementing. Then /clear.
4. Implement — execute the plan phase by phase
/sdw:implement-plan @.claude/ai/plan-YYYY-MM-DD-topic.md
Works through each phase, runs tests to verify, checks off tasks. Reset context between phases.
5. Review — catch issues before committing
/sdw:code-review
Runs tests, checks for security issues, quality problems, test shortcuts. Outputs review.md.
You might wonder: why not run the whole thing automatically in one command?
The pauses are the lesson. At each step you have to look at what Claude produced and make a conscious decision:
Those decisions are the habit this workflow is building. If you skip them, you're not using spec-driven development — you're just generating code faster and hoping it's right.
Mistakes in research compound. If research misses something important and you go straight to implementation, Claude will build the wrong thing confidently. The human checkpoint between each phase exists to catch drift before it becomes expensive.
Context is a resource, not a given. The /clear between phases isn't just housekeeping — it's a forcing function. It makes the document the handoff, not the conversation. research.md is what carries knowledge from phase 1 into phase 2. plan.md carries it from phase 2 into phase 3. The documents are the contract.
The workflow only works if you engage with the outputs. Run the command, read the document, decide to proceed. That's the loop. Once that habit is internalized, you can move faster — but speed comes after the mental model is built, not before.
| Agent | Purpose |
|---|---|
codebase-locator | Finds WHERE code lives — files, directories, entry points |
codebase-analyzer | Explains HOW code works — data flow, logic, patterns |
codebase-pattern-finder | Shows existing patterns with code examples to model after |
test-runner | Discovers and runs the test suite, reports full error details |
code-reviewer | Senior code review with zero tolerance for test shortcuts |
web-search-researcher | Researches web sources for answers to technical questions |
Claude invokes these automatically. You don't call them directly.
| Command | What it does |
|---|---|
/sdw:research-codebase | Explores the codebase and outputs research.md |
/sdw:create-plan | Interactive planning process, outputs plan.md |
/sdw:implement-plan | Implements an approved plan phase by phase |
/sdw:code-review | Reviews recent changes, runs tests, outputs review.md |
/sdw:research-prompt | Transforms a "do X" prompt into a research-focused prompt |
/context to check how much context is used/clear between phases to reset context.claude/ai/ — reference them with @.claude/ai/filename.mdThe plugin gives you a base. To tweak a command or agent for your specific project:
~/.claude/plugins/cache/sdw*/.claude/commands/ or .claude/agents/ in your project/research-codebase (no namespace), the plugin original stays at /sdw:research-codebaseNote: Do not edit files in
~/.claude/plugins/cache/directly. Changes there are wiped on/plugin update.
npx claudepluginhub visma-tech-lt/ai-workshop-marketplace --plugin sdwCode transformation: Dev SDLC orchestrator (code-shipping pipeline), plan, assert, audit, review, test, refactor, debug, for-sure. Hosts engineering agents.
A Claude Code plugin for structured, context-efficient software development.
Verification-first engineering toolkit for Claude Code. 15 skills across a 5-phase spine (Investigate → Design → Implement → Verify → Ship), 8 specialist agents, an interactive setup wizard. Every skill has rationalizations + evidence requirements. Built for senior ICs and tech leads.
Codebase exploration, refactoring, and quality analysis
Persona-driven AI development team: orchestrator, team agents, review agents, skills, slash commands, and advisory hooks for Claude Code
AI-powered development workflow automation - Phase-based planning, implementation orchestration, preflight code quality checks with security scanning, ship-it workflow, and development principles generator for CLAUDE.md