TL;DR
A community project called oh-my-claude-sisyphus packages a multi-agent orchestration system for Claude Code, reworking an earlier oh-my-opencode effort. The tool installs eleven specialized agents, numerous slash commands and lifecycle hooks, and offers automatic updates and composable skills to manage complex developer workflows.
What happened
Oh-my-claude-sisyphus is a ported multi-agent orchestration package that integrates with the Claude Code client. The project describes itself as a resurrection of an earlier package (oh-my-opencode) and provides an installer that populates a ~/.claude directory with agent definitions, command files and a system prompt. Eleven agents cover roles from architecture debugging and documentation research to orchestration and execution. The distribution also registers a set of slash commands (for example /sisyphus, /ultrawork, /deepsearch) and includes a hooks system with 18 lifecycle hooks for context recovery, validation, notifications and automation. The package advertises built-in skills (sisyphus, orchestrator, ultrawork, ralph-loop, frontend-ui-ux, git-master), an intelligent skill-composition feature introduced in v1.8.0, a silent auto-update mechanism, and an npm-friendly SDK entry point for programmatic use.
Why it matters
- Adds a structured multi-agent workflow layer to Claude Code, enabling delegation of subtasks to specialized agents.
- Provides developer-facing automation (commands, hooks, skills) that can standardize repetitive project tasks and documentation generation.
- Built-in auto-update and lifecycle hooks aim to reduce maintenance friction and recover from session or token-limit errors.
- Composable skills and task-driven skill selection could change how complex, multi-step developer tasks are orchestrated inside Claude Code.
Key facts
- Repository name: oh-my-claude-sisyphus (ported from oh-my-opencode).
- Installer populates ~/.claude/ with agents/, commands/ and a CLAUDE.md system prompt file.
- Eleven agents are provided, covering execution, planning, review and orchestration roles (examples: oracle, librarian, explore, prometheus, orchestrator-sisyphus, sisyphus-junior).
- Slash commands include /sisyphus, /sisyphus-default, /ultrawork, /deepsearch, /analyze, /plan, /review, /prometheus, /orchestrator, /ralph-loop and others.
- The package exposes 18 lifecycle hooks for behavior such as rules injection, keyword detection, session recovery, token-limit handling, and background notifications.
- Six builtin skills are listed (sisyphus, orchestrator, ultrawork, ralph-loop, frontend-ui-ux, git-master) and can be composed automatically since v1.8.0.
- Magic keywords (e.g., ultrawork, search, analyze) trigger enhanced behaviors when included in prompts.
- Installer options: one-liner curl for macOS/Linux, npm install -g (all platforms, recommended for Windows), or manual git clone and script execution.
- Auto-update runs silently in the background; a /update slash command is provided to check for updates manually.
- An SDK integration example is shown: npm install oh-my-claude-sisyphus @anthropic-ai/claude-agent-sdk and an exported createSisyphusSession function.
What to watch next
- How widely the package is adopted by Claude Code users and projects (not confirmed in the source).
- Any platform-level policy or compatibility responses from Claude/Anthropic regarding third-party agent orchestrators (not confirmed in the source).
- Further releases and changes to the hook system, skill composition rules and agent roster in future versions.
Quick glossary
- Multi-agent orchestration: Coordinating multiple specialized software agents to divide, delegate and execute subtasks toward a larger goal.
- Claude Code: A client environment for interacting with Claude agents; the project integrates into its ~/.claude configuration directory.
- Slash command: A prefixed command (e.g., /sisyphus) entered into a chat or CLI that triggers a predefined behavior or workflow.
- Lifecycle hook: A programmable event point that runs custom logic at particular stages of a tool's operation, such as recovery or validation steps.
- Skill composition: Combining multiple capability modules so the system applies several complementary behaviors to a single task.
Reader FAQ
How do I install oh-my-claude-sisyphus?
Options include a curl one-liner for macOS/Linux, npm install -g for all platforms (required for Windows), or a manual git clone and script run.
How many agents does it add and what do they do?
The package installs eleven agents with roles across execution, planning, review and orchestration (for example oracle, librarian, explore, prometheus, orchestrator-sisyphus, sisyphus-junior).
Does it update itself?
Yes — the project includes a silent auto-update mechanism plus a /update command to check for updates manually.
Is this an official Anthropic product?
Not confirmed in the source.
Can I customize agents for a specific project?
Yes — agent files in ~/.claude/agents/ can be edited to change descriptions, tools and model selections.
Multi-agent orchestration system for Claude Code Like Sisyphus, these agents persist until every task is complete. Install • Usage • Agents • Architecture • Website The Saga Day 0: oh-my-opencode…
Sources
- Sisyphus Now Lives in Oh My Claude
- code-yeongyu/oh-my-opencode: The Best Agent Harness. …
- Claude Code's Hidden Multi-Agent Orchestration now …
- Multi-Agent Orchestration: Running 10+ Claude Instances …
Related posts
- VAM Seek — 2D thumbnail seek grid for HTML5 video, ~15KB client JS
- A Year of Work on the ALPM Project: 15 Months of Rust-based Package Tools
- Kodbox: Open-source web desktop with multi-cloud storage and web IDE