/spec-to-ralph:generate

Generate an optimized Ralph Wiggum prompt from your Spec Kit project.

Prerequisites

This command supports two Spec Kit structures:

Flat Structure (simple projects)

.specify/
├── constitution.md
├── spec.md
├── plan.md
└── tasks.md

Feature-Based Structure (recommended for larger projects)

.specify/
├── constitution.md
└── features/
    ├── 001-user-auth/
    │   ├── spec.md
    │   ├── plan.md
    │   └── tasks.md
    ├── 002-payment-flow/
    │   ├── spec.md
    │   ├── plan.md
    │   └── tasks.md
    └── 003-dashboard/
        ├── spec.md
        ├── plan.md
        └── tasks.md

Feature Selection

All Features

/spec-to-ralph:generate --feature all

Generates a prompt that works through all features sequentially.

Single Feature

/spec-to-ralph:generate --feature 001
/spec-to-ralph:generate --feature user-auth
/spec-to-ralph:generate --feature 001-user-auth

Any of these formats work - matches by number prefix or name.

Multiple Features

/spec-to-ralph:generate --feature 001,002,003
/spec-to-ralph:generate --feature user-auth,payment-flow

Comma-separated list, processed in the order specified.

Auto-Detection

/spec-to-ralph:generate

• If flat structure: uses root spec/plan/tasks
• If feature structure with one feature: uses that feature
• If feature structure with multiple: prompts for selection

What This Command Does

1. Validates your Spec Kit project structure
2. Extracts constraints from constitution.md
3. Detects your tech stack and appropriate test/lint commands
4. Counts tasks and calculates optimal iteration budget
5. Generates PROMPT.md optimized for Ralph convergence
6. Generates ralph-config.md with the recommended command

Process

Step 1: Detect Project Structure

Check for Spec Kit layout:

Flat structure (if .specify/tasks.md exists):

.specify/
├── constitution.md
├── spec.md
├── plan.md
└── tasks.md

Feature structure (if .specify/features/ exists):

.specify/
├── constitution.md
└── features/
    ├── 001-user-auth/
    ├── 002-payment-flow/
    └── 003-dashboard/

Step 2: Resolve Feature Selection

If --feature argument provided:

• all → Include all features in order (001, 002, 003...)
• Single value → Match by prefix (001) or name (user-auth) or full (001-user-auth)
• Comma-separated → Parse and validate each, maintain specified order

If no --feature argument:

• Flat structure → Use root spec files
• Single feature → Use that feature automatically
• Multiple features → List available and ask user to select

Step 3: Load Feature Files

For each selected feature, load:

• features/NNN-name/spec.md
• features/NNN-name/plan.md
• features/NNN-name/tasks.md

Always load root constitution.md (applies to all features).

Step 4: Analyze Content

Read each file and extract:

From constitution.md (global):

• Non-negotiable rules (bullet points, numbered lists)
• Tech stack requirements
• Testing requirements
• Coding standards

From each feature's plan.md:

• Architecture decisions
• Dependencies
• Build/test commands

From each feature's tasks.md:

• Count of incomplete tasks (- [ ] checkboxes)
• Task descriptions for complexity estimation

Step 5: Detect Tech Stack

Based on constitution and plan content, identify:

Stack	Test Command	Lint Command	Build Command
Node/TS	npm test	npm run lint	npm run build
Python	pytest	ruff check .	-
Rust	cargo test	cargo clippy	cargo build
Go	go test ./...	go vet ./...	go build ./...

If commands are explicitly mentioned in constitution/plan, prefer those.

Step 6: Calculate Iterations

total_tasks = sum of incomplete tasks across all selected features
base_iterations = total_tasks × 4
with_buffer = base_iterations × 1.2
max_iterations = min(with_buffer, 50)  # Cap for cost control

For multi-feature runs, add 5 iterations per feature transition.

Step 7: Generate PROMPT.md

For single feature, generate focused prompt:

# Ralph Loop: [Project] - Feature [NNN-name]

## Context
- `@.specify/constitution.md` - Global rules
- `@.specify/features/NNN-name/spec.md` - This feature's spec
- `@.specify/features/NNN-name/plan.md` - This feature's plan

## Your Mission
Complete all tasks in `.specify/features/NNN-name/tasks.md`

For multiple features, generate sequential prompt:

# Ralph Loop: [Project] - Features [list]

## Context
- `@.specify/constitution.md` - Global rules
- Feature specs in `.specify/features/`

## Your Mission
Complete features in this order:
1. 001-user-auth (8 tasks)
2. 002-payment-flow (12 tasks)
3. 003-dashboard (6 tasks)

## Process
Work through each feature completely before moving to the next.
For each feature:
1. Read its spec.md and plan.md
2. Complete all tasks in its tasks.md
3. When all tasks marked [x], move to next feature

Step 8: Generate ralph-config.md

Create a config file with:

• Recommended /ralph-loop command
• Features included and task counts
• Rationale for iteration count
• Pre-run checklist
• Monitoring tips
• Recovery commands

Step 9: Report Results

Output:

• Files generated
• Features included
• Total task count
• Recommended iterations
• Next step command

Example Usage

Basic (auto-detect):

/spec-to-ralph:generate

All features:

/spec-to-ralph:generate --feature all

Single feature:

/spec-to-ralph:generate --feature 001
/spec-to-ralph:generate --feature user-auth

Multiple specific features:

/spec-to-ralph:generate --feature 001,003
/spec-to-ralph:generate --feature user-auth,dashboard

With iteration override:

/spec-to-ralph:generate --feature all --max-iterations 100

Output

After running, you'll have:

project/
├── PROMPT.md           # Ralph-optimized prompt
├── ralph-config.md     # Recommended settings
└── .specify/
    ├── constitution.md
    └── features/
        ├── 001-user-auth/
        ├── 002-payment-flow/
        └── 003-dashboard/

Next Steps

After generation, run:

/ralph-loop "Follow PROMPT.md to complete all tasks" --max-iterations [N] --completion-promise "ALL_TASKS_COMPLETE"

Or use /spec-to-ralph:start to generate AND start Ralph in one command.