Command Reference
Complete guide to all Spec Kit slash commands.
š Table of Contents
- Command Naming Convention
- Core Commands
- Workflow Comparison
- Command Tips
- Advanced Usage
- What's Next?
Command Naming Convention
All commands use the speckit. prefix to prevent conflicts:
- Type full name:
/speckit.specify - Or just keyword:
/specify(also works!) - Prevents conflicts with other editor slash commands
Core Commands
/speckit.constitution
Purpose: Create or update project constitution (non-negotiable principles)
When to use: First step - establish guardrails
Example:
/speckit.constitution
Static website with minimal dependencies,
optimized for mobile and desktop (responsive),
with accessibility standards (WCAG AA).
What it does:
- Creates
.speckit/memory/constitution.md - Updates dependent templates
- Embeds constitutional checks throughout workflow
Pro tip: Share constitutions across teams for consistency
/speckit.specify
Purpose: Define requirements (WHAT and WHY)
When to use: After constitution - describe what you want
Example:
/speckit.specify
I am building a modern podcast website that should:
- Look sleek and professional
- Have a landing page with one featured episode
- Include episodes page, about page, and FAQ page
- Display 20 mock episodes
- Be mobile responsive
What it creates:
spec.mdwith functional requirements- User stories and acceptance criteria
- Acceptance checklist
Focus on: User needs, requirements, edge cases Avoid: Technology choices, implementation details
/speckit.clarify āØ
Purpose: Interactive Q&A to fix under-specification
When to use: After specify - eliminate assumptions
Example:
/speckit.clarify
AI: What episode metadata should be displayed?
A) Basic (title, date)
B) Standard (title, date, duration, description)
C) Extended (all fields + host, guests, tags)
You: C
What it does:
- Asks up to 5 clarification questions
- Provides multiple choice options
- Auto-updates spec with answers
- Logs session with timestamp
Why use it: Catches blind spots you didn't question
Pro tip: Just type the letter (A/B/C) - super fast!
/speckit.checklist ⨠NEW!
Purpose: Generate domain-specific quality checklists ("unit tests for English")
When to use: After specify/clarify - validate completeness
Syntax: /speckit.checklist <domain>
Examples:
/speckit.checklist UX
/speckit.checklist security
/speckit.checklist accessibility
/speckit.checklist performance
What it generates:
## Requirement Clarity
ā Visual hierarchy defined for all page types
ā Loading states defined for page transitions
ā Error handling specified
## Requirement Consistency
ā Navigation patterns consistent
ā Layout approach aligned
## Testability
ā Metrics include measurement conditions
How to use:
- Generate checklist:
/speckit.checklist UX - Review items
- Cross-check: "Cross-check the UX checklist with spec.md"
- AI updates spec to fill gaps
Domains:
- UX: Visual hierarchy, loading states, error handling, responsive behaviors
- Security: Authentication, authorization, data validation, encryption
- Accessibility: WCAG compliance, keyboard navigation, screen readers
- Performance: Load times, bundle sizes, Core Web Vitals
Why use it: Finds requirements you didn't know you were missing
/speckit.plan
Purpose: Technical planning (HOW to build it)
When to use: After spec validation
Example:
/speckit.plan
Use Next.js with static site configuration.
No databases needed - use mock data.
Make sure the site is responsive and mobile-ready.
What it creates:
plan.mdwith technical details- Framework and language choices
- Architecture decisions
- Dependencies
- Data models
- Testing strategy (optional)
Must respect: Constitution requirements
/speckit.tasks
Purpose: Break down into actionable tasks
When to use: After plan is complete
Example:
/speckit.tasks
Break this down into manageable tasks.
What it creates (NEW format):
## Phase 1: Setup & Shared Infrastructure
- Project scaffolding
- Base configuration
## Phase 2: Foundational Prerequisites
- Core dependencies
- Data structures
## Phase 3: MVP Implementation
- User Story 1: [Core feature]
- Task breakdown
## Phase 4+: Iterations & Improvements
- User Story 2: [Enhancements]
Key improvements:
- ⨠MVP-first approach
- ⨠Phase-based structure
- ⨠User story organization
- ⨠Tests optional (no longer forced!)
To include tests: Explicitly state in plan
/speckit.analyze ā ļø
Purpose: Validate spec, plan, and tasks against constitution
When to use: BEFORE implement - catches inconsistencies
Example:
/speckit.analyze
What it does:
- Analyzes 3 core artifacts (spec, plan, tasks)
- Validates against constitution
- Finds discrepancies by severity:
- Critical: Constitutional violations
- Medium: Structural mismatches
- Low: Minor inconsistencies
- Provides fix recommendations
Example output:
Severity: CRITICAL
Issue: Next.js violates "static HTML/CSS only" constitution
Recommendation: Either justify violation or return to vanilla approach
Severity: MEDIUM
Issue: Plan uses assets/ but tasks use app/ Next.js structure
Recommendation: Update plan to reflect actual structure
Why critical: Spots issues BEFORE coding starts
Best practice: ALWAYS run before /speckit.implement
/speckit.implement
Purpose: Execute implementation
When to use: After analysis passes - build it!
Example:
/speckit.implement
What it does:
- Checks prerequisites (ensures artifacts exist)
- Reads tasks and validates readiness
- Executes tasks sequentially by phase
- Updates task status as it progresses
- Creates all necessary files
- Runs tests (if specified)
Auto-approved scripts (VS Code + Copilot):
- No permission prompts!
- Faster iteration
- Scripts automatically approved
Pro tip: Use Claude* Sonnet 4 or GPT*-4 for best results
Workflow Comparison
Minimum (Quick Prototypes)
1. /speckit.constitution
2. /speckit.specify
3. /speckit.plan
4. /speckit.tasks
5. /speckit.implement
Recommended (Quality Projects)
1. /speckit.constitution
2. /speckit.specify
3. /speckit.clarify ā Eliminates assumptions
4. /speckit.checklist UX ā Validates completeness
5. /speckit.plan
6. /speckit.tasks
7. /speckit.analyze ā Catches inconsistencies
8. /speckit.implement
Comprehensive (Production)
1. /speckit.constitution
2. /speckit.specify
3. /speckit.clarify
4. /speckit.checklist UX
5. /speckit.checklist security
6. /speckit.checklist accessibility
7. Cross-check all checklists with spec
8. /speckit.plan
9. /speckit.tasks
10. /speckit.analyze
11. /speckit.implement
Command Tips
Constitution
ā DO: Encode critical, non-negotiable requirements ā DO: Share across teams for consistency ā DO: Include testing standards, accessibility, security ā DON'T: Put implementation details here
Specify
ā DO: Be specific and detailed ā DO: Include edge cases and acceptance criteria ā DO: Focus on user needs and motivation ā DON'T: Mention technology or frameworks ā DON'T: Include implementation details
Clarify
ā DO: Answer all questions thoughtfully ā DO: Choose specific options (not vague ones) ā DO: Run after initial spec ā DON'T: Skip this step (catches blind spots!)
Checklist
ā DO: Generate multiple domain checklists (UX, security, accessibility) ā DO: Cross-check with spec after generating ā DO: Let AI update spec to fill gaps ā DON'T: Skip cross-checking (defeats the purpose!)
Plan
ā DO: Reference constitution requirements ā DO: Specify all technical choices ā DO: State if tests are NOT needed ā DON'T: Contradict constitution
Tasks
ā DO: Review generated tasks ā DO: Regenerate if spec/plan changes ā DON'T: Edit tasks manually (regenerate instead)
Analyze
ā DO: ALWAYS run before implement ā DO: Address critical issues before proceeding ā DO: Update spec/plan/tasks based on findings ā DON'T: Skip this step (catches major issues!)
Implement
ā DO: Ensure all prior steps completed ā DO: Review code as it's generated ā DO: Commit frequently ā DON'T: Run without analyzing first
Advanced Usage
For Existing Projects
Add these to your commands for better results:
/speckit.constitution
Use existing architecture in [describe folder structure].
Do not introduce new frameworks.
Follow patterns in [specific files/folders].
/speckit.specify
Add feature to existing project.
Must integrate with existing [describe components].
DO NOT modify [list files that shouldn't change].
Encoding Learnings
After implementation:
Based on the conversation, encode the learnings
and experience pieces (not technical details)
into spec.md for the current feature.
Then:
Transform the learnings into relevant functional requirements.
Unwinding Mistakes
If AI goes off-rails:
1. Update spec with correct requirements
2. Commit spec changes
3. Discard code changes in git client
4. Re-implement from updated spec
What's Next?
- New Project Guide - Step-by-step for new projects
- Existing Project Guide - Adding features to existing code
- Complete Guide - In-depth documentation
Master these commands and transform your development workflow! š