Execute an autonomous development loop that picks one task per iteration, implements it, verifies it, and commits the result — each iteration in a fresh context window. Use when user runs /ralph, mentions "ralph loop", "autonomous loop", "builder verifier", "run tasks automatically", "iterate on tasks", "develop autonomously", or wants an automated build-verify-commit cycle with task tracking.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Autonomous development with fresh context per iteration and Hat-lite builder/verifier workflow.
# From project directory (after /create-prd)
./ralph.sh 20 # Run up to 20 iterations
./ralph.sh 5 # Quick test with 5 iterations
Each iteration runs in a fresh context window:
┌─────────────────────────────────────────────────────┐
│ ITERATION (Fresh Context) │
├─────────────────────────────────────────────────────┤
│ 1. ORIENT: Read activity.log + tasks.json + memories│
│ 2. BUILD: Pick ONE task, implement, verify │
│ 3. VERIFY: Review, update status, commit │
│ 4. LEARN: (Optional) Save insights to memories │
│ 5. DECIDE: All done? → RALPH_COMPLETE │
└─────────────────────────────────────────────────────┘
| Role | When | Responsibility |
|---|---|---|
| Builder | ~65% of iteration | Implement ONE task |
| Verifier | ~35% of iteration | Review, update, commit |
| File | Purpose |
|---|---|
.ralph/current-taskset/tasks.json | Task state (source of truth) |
.ralph/current-taskset/activity.log | Iteration history |
.ralph/current-taskset/memories.md | Persistent learnings |
.ralph/current-taskset | Symlink to active task set |
PROMPT.md | Instructions per iteration |
ralph.sh | Loop runner script |
Tasks are stored in .ralph/current-taskset/tasks.json:
{
"tasks": [
{
"id": "setup-001",
"description": "Initialize project",
"verificationTier": "build",
"passes": false,
"iteration_completed": null
}
]
}
Valid verificationTier values: "build" (default), "visual", "api", "e2e". See WORKFLOW.md for details.
When all tasks pass, output:
RALPH_COMPLETE: All tasks verified
Subcommands: /ralph run [N], /ralph status, /ralph init, /ralph taskset <new|list|switch|delete>, /ralph add-task, /ralph remember, /ralph memories.
See WORKFLOW.md for the 5-phase iteration lifecycle (Orient, Build, Verify, Learn, Decide) and Hat-Lite role definitions.
See EXAMPLES.md for web app, API, resumption, and mid-loop status checking scenarios.
See TROUBLESHOOTING.md for immediate exit, premature completion, infinite loop, and context issues.
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Create, review, and manage Architecture Decision Records (ADRs) with auto-numbering, template detection, quality review, and index maintenance. Use when user mentions "ADR", "architecture decision", "document this decision", "create ADR", "review ADR", editing ADR files (docs/adr/, doc/adr/, .adr/), or discussing architectural choices and tradeoffs.
日本語の概要は準備中です。原文の説明を表示しています。
Analyze system architecture, module structure, API contracts, data models, and code patterns. Use when designing systems, reviewing module boundaries, evaluating API designs, analyzing data models, checking pattern conformance, tracing decisions, or reviewing frontend architecture. Triggers: "architecture", "module design", "API design", "data model", "boundaries", "patterns", "decisions", "frontend architecture".
日本語の概要は準備中です。原文の説明を表示しています。
Converts natural-language descriptions or UI spec files into optimized Google Stitch prompts. Use when creating, refining, or validating design directives for Google Stitch. Use when user says "create a Stitch prompt", "optimize this for Stitch", "convert this spec to a Stitch prompt", "write a UI prompt", or mentions Google Stitch prompt authoring. Follows Stitch best practices with short, directive prompts focused on screens, structure, and visual hierarchy.
日本語の概要は準備中です。原文の説明を表示しています。
Zero-dependency live-preview server providing an agent↔browser bidirectional companion. Watches a content directory, wraps HTML fragments in a frame template, broadcasts live-reload over WebSocket, and captures browser-side click events to a JSONL file. Use when an agent needs to show HTML to the user in a browser and react to clicks — mockups, design exploration, gallery browsing, multi-select choices, prototype iteration. Triggered by "live preview", "browser companion", "preview server", "browser mockup", "show in browser", "arkhe-preview", or any skill that wants to write HTML fragments and have the user see them reload in real time.
日本語の概要は準備中です。原文の説明を表示しています。
Deletes merged git branches (local and remote) and flags stale unmerged branches for manual review. Use when user mentions "cleanup branches", "delete merged branches", "prune old branches", "remove stale branches", "branch cleanup", or runs /cleanup-branches command.
日本語の概要は準備中です。原文の説明を表示しています。
Interactive Claude Code environment setup wizard. Detects existing configuration, guides through best-practice setup for Global CLAUDE.md, project scaffolding, MCP servers, hooks, custom agents, keybindings, and settings. Use when user runs /devtools:code-env-setup, mentions "setup claude code", "configure claude", "claude code setup", "environment setup", or "initialize claude code".
日本語の概要は準備中です。原文の説明を表示しています。