Ralph Hitl

Ralph HITL (Human-In-The-Loop) Mode

You are running a single iteration of Ralph with full visibility. This is for learning Ralph's behavior and refining your prompt before going AFK.

Why HITL First?

• Learn how Ralph operates before trusting it with autonomous mode
• Refine your prompt based on what you observe
• Build confidence in the system
• Catch issues early before they compound in AFK mode

HITL vs AFK:

• HITL (this mode): You watch, intervene when needed, learn the system
• AFK (autonomous): Set max iterations, walk away, review commits later

Your Process

Work through ONE task from prd.json, showing me everything:

Step 1: Review Progress & PRD (Show Me)

First, read progress.txt to understand what's already been done. This is how Ralph avoids expensive re-exploration.

Then:

• Read prd.json
• Tell me which task you selected and WHY (explain your priority reasoning)
• Show me the task's acceptance criteria
• Ask if I agree with the choice

Priority Order: architectural > integration > spike/unknown > functional > polish

• Explain why you chose this task over others
• Check dependencies are met

Step 2: Explore (Show Me)

• Tell me which files you'll read
• Show me what patterns you're following from existing code
• Ask if I want you to consider anything else

Step 3: Implement (Show Me)

• Make small, focused changes - one logical thing at a time
• Write the code with visibility
• Explain your decisions as you go
• Show me exactly what you're changing

Step 4: Validate - ALL Feedback Loops (Show Me)

Run ALL feedback loops before committing:

• npm run type-check - Must pass with no errors
• npm run lint - Must pass with zero warnings
• npm run test - All tests must pass
• npm run build - Production build must succeed

DO NOT commit if any fail. Show me the errors and ask how to proceed.

Step 5: Propose Commit (Wait for Approval)

• Show me the exact commit message
• Show me the files that changed (git diff)
• WAIT for my approval before committing

Commit format:

[ralph] feat-XXX: Brief description

- Change 1
- Change 2

PRD: feat-XXX | Iteration: N

Step 6: Update State

• Update prd.json to mark passes: true for this task
• Append to progress.txt with:
  • Task completed and PRD item reference
  • Key decisions made and reasoning
  • Any blockers or notes for next iteration
• STOP here - do NOT continue to next task

Important

• Do NOT loop - this is ONE iteration only for learning
• Show everything - I want to see your thought process
• Wait for approval - don't commit without my say-so
• Answer questions - explain why you're doing things
• Quality over speed - small steps compound into big progress

What to Watch For

While observing Ralph, note:

• Does it pick the right task priority?
• Are the commits focused or too large?
• Do feedback loops catch issues?
• Is the code quality what you expect?

Use these observations to refine your prompt for AFK mode.

Transitioning to AFK

Once you're comfortable with Ralph's behavior:

1. Refine your prompt based on HITL observations
2. Set a modest max iterations (10-20) for first AFK run
3. Review commits when you return
4. Iterate on your prompt

Begin by reading progress.txt (if it exists) and prd.json, then show me what tasks are available.