yeachan-heo/oh-my-claudecode
9 skills5.1k installs324k starsGitHub
Install
npx skills add https://github.com/yeachan-heo/oh-my-claudecodeSkills in this repo
1HudHUD is an oh-my-claudecode configuration skill that tunes the Heads-Up Display rendered on Claude Code’s statusline. Solo builders running OMC can invoke it to repair a broken HUD, pick a density preset, or inspect what is currently enabled without hand-editing JSON and wrapper scripts blind. Setup flow checks for omc-hud.mjs and statusLine configuration, creates missing artifacts, and prompts for a Claude Code restart when needed. Arguments like minimal, focused, or full map to display layouts so you can reduce noise during deep work or surface more context during orchestration sessions. It is narrowly scoped: it writes under your Claude config directory only and documents that scope explicitly. Use early in a Build session when adopting oh-my-claudecode or when the statusline disappears after an upgrade.666installs2RalplanRalplan is the oh-my-claudecode consensus planning entrypoint for solo builders who want agent execution paused until a multi-role plan is deliberate enough to trust. It maps to plan --consensus: Planner, Architect, and Critic iterate with RALPLAN-DR deliberation (short by default), surfacing a final plan marked pending approval unless you pass --interactive for explicit sign-off. High-risk signals—auth, migrations, destructive edits, incidents, compliance—can force deliberate mode with three-scenario pre-mortem and richer test dimensions. Flags let you route the Architect pass to Codex when installed. Without --interactive, the workflow stops after outputting the plan rather than mutating the repo, which makes ralplan ideal as a mandatory front door before ralph/autopilot-style runs that would otherwise steamroll ambiguous requirements.630installs3UltraqaUltraQA is an advanced agent skill from the oh-my-claudecode stack that keeps coding agents in a tight quality loop until explicit gates succeed. After the intended behavior is known, it alternates automated checking (tests, builds, lint, typecheck, or custom patterns) with verification and fixes rather than stopping at a single failed run. Solo builders shipping with Claude Code benefit when they want hands-off cycling comparable to a diligent QA plus architect review, especially on small teams without dedicated QA. The skill documents how it relates to /goal evaluators, Ralph, Team, and Ultragoal so it acts as a sub-loop instead of fighting other session modes. Use it in Ship for testing discipline and during late Build when integrations need repeated validation. It is not for discovering what to build—that remains upstream planning and implementation skills.618installs4DeepinitDeep Init is an agent skill that walks your coding agent through creating layered AGENTS.md documentation from the repository root down through major src and docs folders. Each file summarizes purpose, key files, and subdirectories while linking to its parent AGENTS.md so agents can climb the tree like a table of contents. Solo builders benefit when inheriting a messy repo, starting a monorepo, or whenever Cursor and Claude Code keep guessing wrong about where UI versus utils live. Use it after clone or before a large refactor so subsequent skills hit the right paths. It is documentation for machines first, but it also helps you sanity-check structure. Expect timestamped generated headers and consistent tables agents can scan quickly.557installs5AutopilotAutopilot is a level-4 oh-my-claudecode agent skill for solo and indie builders who want hands-off execution from a terse product description to shipped-quality code. Instead of ad-hoc chat sessions, it runs a structured pipeline: requirements analysis, technical design, implementation planning, parallel coding, QA loops, and multi-perspective validation. Install it when you are willing to let the system run to completion and your task naturally spans planning, building, and testing—not when you only need brainstorming, a draft explanation, a single surgical edit, or plan review. It complements the plan skill (explore and design before commit) and ralph (focused execution) by owning the full stack in one invocation. Expect working artifacts and verification steps rather than options lists or conversational suggestions alone.552installs6RalphRalph is an oh-my-claudecode agent skill that treats task completion as a verified state machine, not a conversational guess. Solo builders use it when a feature or fix must run to completion across multiple agent iterations: stories live in prd.json, each needs passes true plus a reviewer gate, and the loop continues until the completion promise is satisfied or max iterations hit. It layers on ultrawork for parallel execution while persisting session context and retrying when prior attempts fail verification. Trigger it when the user says ralph, don’t stop, must complete, or finish this—avoid it for open-ended exploration (use plan), full idea-to-code autopilot, or trivial one-shot patches. Advanced users configure critic personas and deslop behavior. Expect tight coupling to PRD artifacts and oh-my-claudecode level-4 orchestration; success means reviewer-signed stories, not partial diffs.542installs7UltraworkUltrawork is a parallel execution protocol for Claude Code-style setups where several subtasks do not depend on each other and can run at once. It grounds intent, gathers context in parallel where possible, and builds dependency-aware graphs when work is not trivially independent. Summaries stay concise and evidence-backed so you can see what delegated agents actually did. It is deliberately not a persistence or verification mode: when you need guaranteed completion with checks, the skill itself points you to ralph; when you need a full autonomous pipeline, it points you to autopilot. Solo builders reach for ultrawork during Build when splitting a feature into parallel research or implementation tracks, during Ship when running concurrent fix passes, or during Operate when triaging unrelated incidents—anytime concurrency beats a single serial executor.535installs8Omc SetupOMC Setup is the mandatory configuration skill for oh-my-claudecode: one command that installs or refreshes the OMC stack whether you came from the Claude marketplace plugin, a global npm package, or a checked-out local worktree. Solo builders invoke it when first enabling OMC or when an update broke hooks, paths, or partial installs. The skill parses CLI-style flags, can scope setup to local or global targets, and supports a forced full pass that skips the pre-setup health check. Documentation stresses immediate execution of the phased workflow (not paraphrasing steps back to the user) and honors CLAUDE_CONFIG_DIR when resolving config directories. It is narrowly aimed at Claude Code power users who want the official repair path instead of hand-editing dotfiles.531installs9Ai Slop CleanerAI Slop Cleaner is an Oh-My-ClaudeCode level-3 skill for solo builders whose working code feels bloated after agent-assisted implementation: duplicate logic, dead paths, needless abstraction layers, boundary leaks, and thin tests. It enforces an OMC execution posture—preserve behavior, prefer deletion over addition, reuse existing utilities, and write a cleanup plan before touching files. When practical, it starts by adding focused regression tests so simplification stays safe. Use it when the user says deslop, anti-slop, or AI slop, or when follow-up work left noisy structure without new product intent. It is not for feature delivery, broad redesigns, or generic refactors without simplification intent, and it should be skipped when behavior is too unclear to verify. The skill fits Ship review on a PR branch and Build when post-agent passes need discipline before you call the task done.460installs