Quick Reference Cheatsheet

The Spec-Driven Workflow

Constitution → /speckit.specify → resolve [NEEDS CLARIFICATION] markers
→ /speckit.plan → /speckit.tasks → Task-by-task: Agent Mode → Code Reviewer → mark done
→ Push to GitHub Issues (MCP) → Generate from Issue

spec-kit Slash Commands

Use these in Copilot Chat in VS Code, not in the terminal.

Command	Purpose
/speckit.specify <feature description>	Create a numbered feature branch and specs/<feature-branch>/spec.md
/speckit.plan <implementation direction>	Generate plan.md plus supporting design docs from the feature spec
/speckit.tasks	Generate tasks.md from plan.md and related design documents

File Locations

File	Path	Purpose
Workspace instructions	.github/copilot-instructions.md	Global project context for all Copilot interactions
File-specific instructions	.github/instructions/*.instructions.md	Context scoped by applyTo glob pattern
Agent	.github/agents/<name>.agent.md	Specialized Copilot persona
Prompt	.github/prompts/<name>.prompt.md	Reusable prompt workflow (invoke from Copilot Chat by name)
MCP config	.vscode/mcp.json	MCP server configuration
Constitution	memory/constitution.md	Project-wide decisions and values
Feature spec	specs/<feature-branch>/spec.md	Feature goal, AC, DoD, clarifications
Plan	specs/<feature-branch>/plan.md	Ordered implementation phases
Tasks	specs/<feature-branch>/tasks.md	Atomic implementation units

Prompt files are covered in L5.3. Unlike agents (which define a persistent persona), prompt files define a one-shot reusable workflow — for example, "run the Code Reviewer against the latest implementation". Create them at .github/prompts/<name>.prompt.md and invoke from Copilot Chat.

Feature Spec Template

# Feature Specification: [Feature Name]

## Goal
[One sentence: what will the user be able to do when this is done]

## Acceptance criteria
- [ ] [Verifiable behavior from the outside]
- [ ] [Another verifiable behavior]

## Out of scope
- [Explicit non-goal 1]
- [Explicit non-goal 2]

## Definition of done
- All acceptance criteria pass
- Unit tests cover [specific functions]
- No TypeScript errors

Task Template

# Task: [Task Name]

## Scope
[One paragraph: what this task implements and nothing more]

## Definition of done
- [ ] [Verifiable criterion 1]
- [ ] [Verifiable criterion 2]

Code Reviewer Agent Prompt Pattern

Review the code just generated for [task name].
Check: Stack compliance, Error handling, Quality.
For each issue: File/Location, Category, Issue, Fix.
If no issues: state what you checked.

MCP Setup (.vscode/mcp.json)

{
  "servers": {
    "github": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "${env:GITHUB_TOKEN}"
      }
    }
  }
}

Set environment variable:

# PowerShell
$env:GITHUB_TOKEN = "ghp_your_token"
# bash
export GITHUB_TOKEN="ghp_your_token"

Failure Category → Practice Map (M1)

Failure category	Practice that fixes it	Module
Requirements gap	Instructions	M2 L2.1
Technical mismatch	Agents	M2 L2.2
Quality/risk issue	Specs	M3

Plan Mode vs. Agent Mode Quick Guide

Use Plan Mode when...	Use Agent Mode when...
Requirement is fuzzy	Requirement is locked
You need to surface assumptions	You need code, not discussion
First encounter with a feature	Implementing from a spec task

Module Navigation

Module	First lesson	Duration
M1	L1.1 The Vibing Trap	~15 min
M2	L2.1 Instructions	~25 min
M3	L3.1 Cost of No Specs	~45 min
M4	L4.1 What Is MCP	~20 min
M5	L5.1 Full Simulation	~35 min