Workflow Setup

Bootstrap a new or existing project with the mycelium workflow structure.

Your Task

Setup Workflow

This workflow guides you through bootstrapping a project with the mycelium workflow structure, detecting project type, gathering configuration through interactive questions, creating the .mycelium/ directory structure, and initializing git if needed.

---

Phase 1: Bootstrap

Step 0: Pre-flight Check

Before starting setup, check if .mycelium/ already exists and assess its completeness.

If .mycelium/ does not exist

→ Continue to Step 1 (fresh setup)

If .mycelium/ exists - Check completeness

Minimum required files:

• .mycelium/state.json
• .mycelium/context/product.md
• .mycelium/context/tech-stack.md
• .mycelium/context/workflow.md
• .mycelium/solutions/patterns/critical-patterns.md

See .mycelium/ directory structure for complete specification.

If all required files present (complete structure):

✅ Mycelium structure already exists and is complete.

Structure:
  ✓ .mycelium/state.json
  ✓ .mycelium/context/product.md
  ✓ .mycelium/context/tech-stack.md
  ✓ .mycelium/context/workflow.md
  ✓ .mycelium/solutions/patterns/critical-patterns.md
  ✓ .mycelium/plans/
  ✓ .mycelium/solutions/

Next steps:
  • Review context files: .mycelium/context/
  • Start planning: /mycelium-plan [description]
  • Full workflow: /mycelium-go [description]

→ EXIT (setup already complete)

If partially complete (missing required files):

⚠️  Incomplete .mycelium/ structure found.

Present:
  ✓ .mycelium/state.json
  ✓ .mycelium/context/product.md

Missing:
  ✗ .mycelium/context/tech-stack.md
  ✗ .mycelium/context/workflow.md
  ✗ .mycelium/solutions/patterns/critical-patterns.md

Options:
  1. Complete the missing files (recommended)
  2. Start fresh (deletes existing .mycelium/)
  3. Exit (manual fix required)

Enter 1-3:

Handle user choice:

• Option 1 (Complete): Continue to Step 4 (Create Directory Structure) to create only missing files
• Option 2 (Start fresh): rm -rf .mycelium/, then continue to Step 1
• Option 3 (Exit): Show guidance on how to manually fix, then EXIT

---

Step 1: Check for Existing Setup

Check if setup_state.json exists and determine next action.

┌─────────────────────────────────────┐
│ Does setup_state.json exist?       │
└─────────┬───────────────────────────┘
          │
    ┌─────┴─────┐
    NO          YES
    │           │
    │     ┌─────┴──────────────────────┐
    │     │ Check status field         │
    │     └─────┬──────────────────────┘
    │           │
    │     ┌─────┴─────┐
    │     │           │
    │  "completed"  "in_progress"
    │     │           │
    │     │           │
┌───▼─────▼───┐   ┌──▼─────────────────┐
│ Path A:     │   │ Path C:            │
│ Start fresh │   │ Resume from        │
│ setup       │   │ checkpoint         │
│             │   │                    │
│ → Step 2    │   │ Show progress      │
│             │   │ → current_section  │
└─────────────┘   └────────────────────┘
                  │
  ┌───────────────┘
  │ Path B:
  │ Setup complete
  │
  │ Info: Already done
  │ → EXIT
  └────────────────

Path A: No setup_state.json

→ Fresh setup, continue to Step 2

Path B: setup_state.json exists with status="completed"

✅ Setup already complete!

Completed: {completed_at}
Files created: {count} files

Next steps:
  /mycelium-plan [description]
  /mycelium-go [description]

→ EXIT

Path C: setup_state.json exists with status="in_progress"

⏸️  Resuming interrupted setup...

Progress:
  ✓ {completed_sections}
  → {current_section} (question {current_question})

Last updated: {last_updated}

Options:

1. Resume - Continue from where you left off
2. Restart - Delete state and start fresh
3. Exit - Manual intervention needed

Ask user choice → handle accordingly:

• Option 1 (Resume): Load answers, continue from current_section and current_question
• Option 2 (Restart): Delete .mycelium/setup_state.json, continue to Step 2
• Option 3 (Exit): Show current state details, EXIT

---

Step 2: Detect Project Type

Analyze the current directory to determine if this is a new or existing project.

Greenfield Detection

Indicators (new project):

• Empty directory or only has README
• No package.json, go.mod, Cargo.toml, requirements.txt, pyproject.toml
• No src/, lib/, app/, pkg/ directories
• No existing source code

Action: Run full interactive setup with all questions

Brownfield Detection

Indicators (existing project):

• Has package.json, go.mod, requirements.txt, Cargo.toml, etc.
• Has existing source code in src/, lib/, app/, etc.
• Has test files
• Has existing git repository

Action:

• Extract context from existing files
• Pre-populate answers by analyzing:
  • package.json → project name, dependencies, scripts
  • README.md → description, purpose
  • Existing code → language, framework, patterns
  • .gitignore → deployment target hints
• Ask user to confirm/correct detected values
• Only ask for information that cannot be inferred

---

Step 3: Interactive Setup

CRITICAL RULES:

1. ✅ Ask ONE question per turn
2. ✅ Wait for user response before proceeding
3. ✅ Offer 2-4 suggested answers + "Type your own" option
4. ✅ Maximum 5 questions per section
5. ✅ Update setup_state.json after EACH successful answer
6. ✅ Validate file write succeeded before continuing

Ask questions ONE AT A TIME following this pattern:

PRODUCT Section (what & why)

Question 1 of 4: Project Identity

What is this project?

Suggested answers:
  1. TaskFlow - A workflow automation tool
  2. DataHub - A data integration platform
  3. CodeReview - An automated code review system
  4. Type your own answer

Enter 1-4:

After user responds:

1. If 1-3: Extract project name and description
2. If 4: Prompt for custom "ProjectName - Description"
3. Save to setup_state.json:

   {
     "current_section": "product",
     "current_question": 2,
     "completed_sections": [],
     "answers": {
       "project_identity": "TaskFlow - A workflow automation tool"
     }
   }
   

4. Verify write succeeded: const state = await readFile('setup_state.json')
5. If write failed: Error "❌ Failed to save answer, cannot continue"
6. Continue to Question 2

Question 2 of 4: Problem Statement

What problem does this project solve?

Suggested answers:
  1. Manual processes waste developer time
  2. Lack of integration between tools
  3. Poor visibility into workflows
  4. Type your own answer

Enter 1-4:

After user responds: Save answer, update current_question: 3, verify, continue

Question 3 of 4: Target Users

Who are the target users?

Suggested answers:
  1. Software development teams (5-50 people)
  2. DevOps engineers
  3. Engineering managers
  4. Type your own answer

Enter 1-4:

After user responds: Save answer, update current_question: 4, verify, continue

After Question 3 (Target Users) complete: Update state:

{
  "current_section": "tech_stack",
  "current_question": 1,
  "completed_sections": ["product"]
}

Note: Goals are optional and can be added later to .mycelium/context/product.md

TECH STACK Section (with what)

For brownfield projects: Detect language/framework from files first, then confirm with user.

Question 1 of 4: Primary Language

What programming language?

Suggested answers:
  1. TypeScript
  2. Python
  3. Go
  4. Type your own (JavaScript, Rust, Java, etc.)

Enter 1-4:

After user responds: Save answer, update current_question: 2, verify, continue

Question 2 of 4: Framework

What framework or library?

Suggested answers:
  1. React + Next.js
  2. FastAPI (Python)
  3. Gin (Go)
  4. Type your own

Enter 1-4:

After Question 2 (Framework) complete: Update state:

{
  "current_section": "workflow",
  "current_question": 1,
  "completed_sections": ["product", "tech_stack"]
}

Note: Database and infrastructure are optional and can be added later to .mycelium/context/tech-stack.md

WORKFLOW Section (how to work)

Question 1 of 1: TDD Strictness

How strict should TDD enforcement be?

Suggested answers:
  1. strict - All code follows RED-GREEN-REFACTOR (recommended)
  2. flexible - TDD for new features, optional for fixes
  3. none - No TDD enforcement
  4. Type your own approach

Enter 1-4:

Default: strict

After user responds:

1. Save answer, update:

   {
     "current_section": null,
     "current_question": null,
     "completed_sections": ["product", "tech_stack", "workflow"],
     "answers": {
       "tdd_strictness": "strict",
       "commit_style": "conventional",
       "review_required": "recommended",
       "coverage_target": 80
     }
   }
   

2. Automatically apply recommended defaults:
   • Commit style: conventional (feat/fix/chore format)
   • Code review: recommended (for major changes)
   • Coverage target: 80%
3. Verify, continue to Step 4 (Create Directory Structure)

Note: Advanced settings (commit style, review requirements, coverage) use recommended defaults. Edit .mycelium/context/workflow.md to customize later.

---

Step 4: Create Directory Structure

Once all answers collected, create this structure:

project/
├── CLAUDE.md                    # Quick context file
├── .mycelium/                   # AI workflow artifacts
│   ├── context/                 # Project information
│   │   ├── product.md
│   │   ├── tech-stack.md
│   │   └── workflow.md
│   ├── plans/                   # Living plan documents
│   ├── solutions/               # Captured learnings
│   │   ├── performance-issues/
│   │   ├── database-issues/
│   │   ├── security-issues/
│   │   ├── ui-bugs/
│   │   ├── integration-issues/
│   │   └── patterns/
│   │       └── critical-patterns.md
│   ├── state.json               # Session state
│   └── setup_state.json         # Setup progress
└── docs/                        # User documentation (if not exists)

See .mycelium/ directory structure for complete documentation of this structure and all file purposes.

File Templates

CLAUDE.md:

# {project_name}

{product_description}

## Quick Context

**What**: {problem_statement}
**Who**: {target_users}
**Stack**: {primary_language} + {frameworks}

## Key Goals

{list_of_goals}

## Getting Started

See `.mycelium/context/` for detailed project information.

## Workflow

- `/mycelium-plan` - Plan new work
- `/mycelium-work` - Execute tasks
- `/mycelium-review` - Review code
- `/mycelium-capture` - Capture learnings

.mycelium/context/product.md:

---
product_name: {name}
description: {description}
target_users: {users}
created: {timestamp}
---

# Product Context

## Problem Statement

{problem}

## Goals

{goals}

## Success Criteria

{how_we_measure_success}

.mycelium/context/tech-stack.md:

---
primary_language: {language}
frameworks: [{frameworks}]
database: {database}
infrastructure: {infrastructure}
updated: {timestamp}
---

# Tech Stack

## Languages

{list}

## Frameworks & Libraries

{list}

## Database

{details}

## Infrastructure

{details}

## Development Tools

{linters, formatters, test frameworks}

.mycelium/context/workflow.md:

---
tdd_strictness: {strict|flexible|none}
commit_style: {conventional|descriptive|atomic}
review_required: {required|recommended|optional}
coverage_target: {percentage}
---

# Workflow Configuration

## TDD Policy

{description_based_on_strictness}

## Commit Guidelines

{description_based_on_style}

## Code Review

{description_based_on_requirement}

## Quality Standards

- Test coverage: ≥{target}%
- No linting errors
- All tests passing

.mycelium/solutions/patterns/critical-patterns.md:

# Critical Patterns

Patterns that MUST be followed in this codebase.

## When to Update This File

After discovering:
- Security-critical patterns
- Performance-critical patterns
- Data integrity patterns
- Architectural constraints

## Patterns

(Will be populated as patterns are discovered)

---

Step 5: Initialize Git (OPTIONAL but RECOMMENDED)

Git is strongly recommended for the mycelium workflow but not required.

With Git Enabled

Benefits:

• ✅ Version control for all changes
• ✅ Parallel implementation via git worktree
• ✅ Automatic commit creation with Co-Author attribution
• ✅ Branch-based feature development
• ✅ Easy rollback of changes
• ✅ Track history of solutions and patterns

Without Git

Limitations:

• ❌ No worktree support (cannot run tasks in parallel)
• ❌ No automatic commits
• ❌ No version control or rollback
• ⚠️ Single-threaded implementation only
• ⚠️ Manual tracking of changes required

Setup still works without git - all core workflow features function, but parallelization and version control are unavailable.

Initialize Git

# Check if git repository exists
if [ ! -d .git ]; then
  echo "Git repository not found."
  echo ""
  echo "Initialize git now? (recommended)"
  echo "  • Enables parallel implementation"
  echo "  • Automatic version control"
  echo "  • Easier rollback"
  echo ""
  # Ask user (use AskUserQuestion tool)
  # If yes:
  git init
  echo "✓ Initialized git repository"
  git checkout -b main
  # Save to setup_state.json: git_enabled: true
  # If no:
  echo "⚠️  Continuing without git (parallel features disabled)"
  # Save to setup_state.json: git_enabled: false
else
  echo "✓ Git repository found"
  # Verify main/master branch exists
  git branch --show-current || git checkout -b main
  # Save to setup_state.json: git_enabled: true
fi

# If git_enabled, create/update .gitignore

Create/Update .gitignore (if git_enabled)

Use language-specific templates from templates/gitignore/:

Detect language and load appropriate template:

# Node.js/TypeScript
if [ -f package.json ]; then
  cat templates/gitignore/node.gitignore >> .gitignore
fi

# Python
if [ -f requirements.txt ] || [ -f pyproject.toml ]; then
  cat templates/gitignore/python.gitignore >> .gitignore
fi

# Go
if [ -f go.mod ]; then
  cat templates/gitignore/go.gitignore >> .gitignore
fi

# Rust
if [ -f Cargo.toml ]; then
  cat templates/gitignore/rust.gitignore >> .gitignore
fi

Always add mycelium-specific entries:

# Mycelium
.worktrees/
.mycelium/state.json

Available templates:

• Node.js gitignore
• Python gitignore
• Go gitignore
• Rust gitignore

Create Initial Commit (if git_enabled)

git add .mycelium/ CLAUDE.md .gitignore
git commit -m "Initialize project with mycelium workflow

- Add .mycelium/ directory structure
- Add CLAUDE.md quick context
- Configure git and .gitignore

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>"

---

Step 6: Import Existing Configs (Brownfield Only)

If project uses linters/formatters:

Detect:

• .eslintrc.* → ESLint
• .prettierrc.* → Prettier
• pyproject.toml → Black, Ruff
• .golangci.yml → golangci-lint
• rustfmt.toml → rustfmt

Action:

• Note their locations in tech-stack.md
• Do NOT create new configs
• Do NOT modify existing configs
• Just document what exists

Example addition to tech-stack.md:

## Code Quality Tools

- ESLint: `.eslintrc.json` (existing)
- Prettier: `.prettierrc` (existing)
- TypeScript: `tsconfig.json` (existing)

---

Step 7: Finalize Setup

Update setup_state.json:

{
  "status": "completed",
  "project_type": "greenfield",
  "completed_at": "2026-02-04T10:30:00Z",
  "files_created": [
    "CLAUDE.md",
    ".mycelium/context/product.md",
    ".mycelium/context/tech-stack.md",
    ".mycelium/context/workflow.md",
    ".mycelium/solutions/patterns/critical-patterns.md",
    ".gitignore"
  ]
}

Initialize state.json:

{
  "project_name": "{name}",
  "initialized_at": "2026-02-04T10:30:00Z",
  "current_track": null,
  "discovered_capabilities": {
    "skills": [],
    "agents": [],
    "patterns": []
  },
  "active_worktrees": [],
  "stats": {
    "tracks_completed": 0,
    "solutions_captured": 0,
    "patterns_discovered": 0
  }
}

Output summary:

✅ Mycelium setup complete!

Created:
  ✓ CLAUDE.md - Quick context
  ✓ .mycelium/context/ - Project information
  ✓ .mycelium/plans/ - Living plans
  ✓ .mycelium/solutions/ - Knowledge base
  ✓ Git repository initialized

Configuration:
  • Language: {language}
  • TDD: {strictness}
  • Coverage: ≥{target}%

Next steps:
  1. Review CLAUDE.md and .mycelium/context/
  2. When you have a task: /mycelium-plan [description]
  3. Or for full automation: /mycelium-go [description]

---

State Management

The setup_state.json enables resume functionality by tracking setup progress.

State structure:

{
  "status": "in_progress",
  "project_type": "greenfield|brownfield",
  "current_section": "product|tech_stack|workflow",
  "current_question": 2,
  "completed_sections": ["product"],
  "answers": {
    "product_name": "MyApp",
    "product_description": "...",
    "problem_statement": "...",
    "target_users": "..."
  },
  "files_created": [".mycelium/context/product.md"],
  "started_at": "2026-02-04T09:00:00Z",
  "last_updated": "2026-02-04T09:15:00Z"
}

Save after EACH answer to enable interruption and resume.

---

Resume Support

When resuming (state shows "in_progress"):

1. Automatically detect in-progress setup from setup_state.json
2. Show progress:

   Resuming setup...
   
   Completed:
     ✓ Product section (4 questions)
   
   In progress:
     • Tech Stack section (question 2 of 4)
   
   Continuing...
   

3. Jump to current_section and current_question
4. Do NOT re-ask answered questions
5. Continue until completion

---

Error Handling

Git init fails:

• Show error message
• Offer to continue without git (with feature limitations)
• If user declines git: set git_enabled=false, continue setup
• If user wants git: suggest resolving git issues, then re-run setup

Directory not writable:

• Check permissions
• Inform user about permission issues
• Suggest solutions

User answers unclear:

• Ask for clarification
• Do NOT assume or guess
• Provide examples of good answers

Interrupted:

• State is saved after each answer in setup_state.json
• Running /mycelium-setup again automatically resumes from checkpoint
• No data loss

---

Greenfield vs Brownfield Differences

Aspect	Greenfield	Brownfield
Questions	All 12 questions	Only undetectable info
Detection	Manual entry	Auto-detect from files
Confirmation	Not needed	Confirm detected values
Git	Create new repo	Use existing
Configs	None created	Import existing

---

Quick Example

/mycelium-setup  # Setup project (auto-resumes if interrupted)

Important Notes

• Git is recommended but optional - Mycelium works without git, but parallel features and version control require it
• One question at a time - Interactive setup asks questions sequentially
• State saved after each answer - Automatically resumes if interrupted
• Brownfield detection - Auto-detects project info from existing files
• Ask ONE question at a time, not all at once
• Save state after EACH answer (enables resume)
• For brownfield: Show detected values, ask confirmation
• Be conversational but concise
• Do NOT proceed to planning - setup only creates structure
• Do NOT create application code - only workflow infrastructure
• Without git: no worktrees, no auto-commits, single-threaded only
• All file paths are absolute when saved to state

---

Next Steps After Setup

1. Review created files (CLAUDE.md, .mycelium/context/)
2. Customize if needed (edit context files)
3. Start work:
   • Manual: /mycelium-plan [task]
   • Autonomous: /mycelium-go [task]

Skills Used

• context: For extracting information from existing projects (brownfield)

References

• .mycelium/ directory structure
• Session state docs
• Session state schema
• Setup state schema
• CLAUDE.md template
• Product template
• Tech stack template
• Workflow template
• Critical patterns template
• Node.js gitignore
• Python gitignore
• Go gitignore
• Rust gitignore
• Enum definitions