ralph-creator（拉尔夫创造者）

Ralph Creator

Generate AFK Ralph loop scripts. Ralph runs claude -p --dangerously-skip-permissions in a loop — one task per iteration, progress tracked between iterations, stops when all tasks are done.

Output Structure

All artifacts go in .ralph/ at the project root:

.ralph/
├── ralph-<name>.sh    # Loop script (executable)
├── backlog.md         # Task checklist
├── progress.md        # Working memory (delete after run)
└── lessons.md         # Mistakes to avoid (persists)

Workflow

1. Gather Context

Ask the user (2-3 questions max, skip if obvious from context):

• What's the task?
• What context files should the agent read?
• Roughly how many subtasks? (determines iteration count)

2. Create .ralph/ Directory

mkdir -p .ralph

3. Write the Backlog

Create .ralph/backlog.md. See references/backlog-format.md.

Rules:

• Each task: - [] checkbox, bolded title, clear description
• One task = one iteration = one context window
• Order by dependency, then priority
• Enough detail for a fresh Claude instance to execute without ambiguity

4. Generate the Script

Create .ralph/ralph-<name>.sh from references/script-template.md.

Customize the placeholders:

• {{ADDITIONAL_CONTEXT_REFS}} — extra @path refs for skills, specs, style guides
• {{ONE_SENTENCE_DIRECTIVE}} — imperative statement: what each iteration produces
• {{EXECUTE_STEPS}} — numbered sub-steps for the domain-specific work
• {{VERIFY_INSTRUCTION}} — how to confirm the work is correct
• {{ADDITIONAL_RULES}} — domain-specific constraints

CRITICAL: The @ refs and prompt text MUST be in ONE quoted string passed to claude -p. The template handles this correctly — do not separate them into different variables or CLI args.

If the task needs a style guide or reference material, put it in a SEPARATE file and add it to {{ADDITIONAL_CONTEXT_REFS}} as another @ reference. Do NOT inline it in the prompt.

5. Create Tracking Files

.ralph/progress.md:

# Progress
Working memory for ralph-<name>. Delete after run.
---

.ralph/lessons.md:

# Lessons
Patterns and mistakes. Review at start of each iteration.
---

See references/tracking-format.md for entry format.

6. Make Executable and Report

chmod +x .ralph/ralph-<name>.sh

Tell the user the run command:

.ralph/ralph-<name>.sh <iterations>

Suggest iterations = task count + 2 (buffer for retries + final COMPLETE check).

Prompt Design Rules

1. Imperative first line — The prompt MUST open with a direct command: YOUR JOB: <one sentence>. The model reads @ context files first (often 1000+ lines), so the prompt must immediately snap it into execution mode, not summary mode.
2. Numbered steps, not section headers — Use STEPS: with numbered items (1. Do X. 2. Do Y.), NOT markdown ## Section headers. Section headers read as documentation. Numbered steps read as commands.
3. Explicit "Do NOT" rules — End the prompt with a RULES: block that includes what the agent must NOT do (e.g., "Do NOT summarize the spec. Do NOT ask what to work on. Just execute."). Without these, the model defaults to helpful-assistant mode.
4. Short and punchy — Prompt should be under 30 lines. The @ files provide all the context. The prompt is ONLY for instructions. If domain-specific guidance (style guides, conventions) is needed, put it in a separate @-referenced file, not inline.
5. @ context refs — Load docs upfront, don't waste tokens exploring
6. Verification step — Agent confirms work before marking done
7. Quality bar — Measurable "done" criteria, no ambiguity
8. One task per iteration — Always include: ONLY work on ONE task. Do not continue to the next.
9. Completion sigil — Always: If all tasks are done, output <promise>COMPLETE</promise>.
10. Tool instructions — If using MCP tools, APIs, or CLIs, include explicit how-to in the execute steps

Prompt Anti-Patterns (Avoid These)

These patterns cause the agent to summarize context instead of executing tasks: