Orca DAG skill
A skill breaks a requirement into an Orca task graph, and a browser viewer shows it and runs workers along the dependencies.
- 144 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The built-in coordinator launches autonomous workers that edit project code.
- The viewer runs on a local port and drives Orca orchestration state.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/orca-dagDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The built-in coordinator launches autonomous workers that edit project code.
- The viewer runs on a local port and drives Orca orchestration state.
This is third-party code. Review the repository files before installing.
What it does
Orca DAG has two independent parts. The `orca-dag` skill teaches your own agent (Claude Code, Codex, Cursor, OpenCode, Gemini CLI and others) to break a requirement into an Orca orchestration task graph and sets the build conventions. The agent stays the planning brain, and there is no embedded SDK. The second part, a viewer from the `orca-dag` package, connects to Orca orchestration state and draws the graph live. Each node picks its own harness (claude, kimi, opencode, grok) and optionally a model. A run button starts the built-in coordinator, which dispatches ready tasks in parallel along the dependencies to autonomous workers until the whole graph is done.
Who it is for. Orca users who run large tasks through several agents and want to see the dependency graph and choose a model for each step.
Good fit when
- A requirement must become a graph of dozens of tasks with dependencies.
- Different nodes need different harnesses: coding in one agent, review in another.
- Tasks run in parallel until the whole graph is done.
- The graph must be redrawn after a task or dependency changes.
Not a fit when
- The project is not managed by Orca or the Orca version is older than 1.4.160.
- You need to edit a single task or dependency in place, which means redrawing the graph.
- A single agent without parallel workers is enough.
Example request
Build an Orca graph for migrating the auth module: schema, service, tests, review. Then open the viewer and wait for me to pick harnesses for the nodes.Limitations
Orca 1.4.160 or newer is required: older versions are unsupported after the orchestration contract rewrite. The official run and coordinator commands in Orca were retired, so the viewer acts as its own coordinator. Orca has no interface to edit one task, so the agent redraws the graph. Installation needs Node.js 20 or newer, and standalone binaries exist. Do not leave a stale coordinator bound to a run, because it keeps feeding your agent.
How to disable. Run `npx orca-dag uninstall`, first with `--dry-run`. It removes the skill from every agent and closes a stale coordinator. Add `--purge` to delete `.orca-dag.config.json`.
FAQ
What does the single install command do?
`npx orca-dag` installs the skill into every agent found (Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Droid and `~/.agents/skills`) and starts the viewer on port 8787.
How do I leave skill folders alone?
Add `--no-skill` or the `ORCA_DAG_NO_SKILL=1` variable. `PORT` sets the port and `NO_OPEN=1` stops the browser from opening.
Is Node.js required?
Node.js 20 or newer is needed. Without it use a standalone binary from the releases page, which only needs `orca` on `PATH`.
What should I check in Orca first?
Run `orca repo add .` for a new repository and `orca status --json`: `runtime.state` should be `ready`, otherwise run `orca open`.
Why does the viewer coordinate itself?
Orca deliberately ships no scheduler and its earlier run commands were retired. The viewer assigns workers and concurrency, as the project README explains.
Discussion
No questions or reviews yet.
Related
SkillsOne agent launches others in the background: Claude Code, Codex, Copilot CLI, Grok Build and Pi, with a model chosen per stage.
A skill that turns the main Claude Code session into a dispatcher: tasks go to background subagents and the chat never blocks.
A process kit for agents: a skill router, project memory in a DEV folder and result verification before saying done.
Skills and the porch CLI for agent orchestration: a strong model plans, cheaper ones code, several model families review.
A task-governance skill for agents: aligns goal, facts and boundaries, takes the shortest path and verifies the result.
A planner for Claude Code, Codex and OpenCode: one goal becomes a task plan and independent tasks run in parallel
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Orca DAG (this card) | no | 144 | |
| Orchestrator | no | 119 | one agent starts others in the background with a model per stage; Orca DAG shows a dependency graph and dispatches tasks per node with a chosen harness |
| Claude Squad | no | 8,790 | runs parallel sessions in tmux with no shared graph; Orca DAG builds a graph and runs execution along dependencies |
README in short
Planning stays in your agent while execution is handled by the viewer. One command installs both parts, and node settings live in `.orca-dag.config.json`.