RadOrchestration

Use this skill once you are aligned on a project and ready to start. Typically this is invoked directly by the user after /rad-brainstorm has completed and the Requirements doc is approved. It is the reference for how to run the planning pipeline and author the Master Plan.

Drive the release flow locally — context, version reconcile, build, validate, CHANGELOG, commit, sync plugins to the marketplace, and tag/push (which triggers the CI npm publish + GitHub Release), then a post-release dev bump.

Use this skill if you are a main agent, coder agent and have this skill assigned. It is the reference for how to perform source-control operations like creating commits, opening pull requests, and creating or cleaning up worktrees for a project.

Use this skill if you're a main agent orchestrator executing a rad-orchestration project. It is the reference for how to run the pipeline, signal events, execute actions, and manage the overall flow of a project. Typically, you are guided to using this skill when /rad-execute is invoked, but you may read it if you need to refresh your memory due to compaction during an execution.

Rebuild the standard installer and reinstall one harness (claude, copilot-vscode, or copilot-cli) locally — build the rad-orc tarball, stop the UI, uninstall then reinstall the harness, restart the UI. Use to dogfood canonical harness-files/ agent and skill edits into ~/.claude or ~/.copilot.

Use this skill if you are a main agent who is helping to author planning documents (Requirements or Master Plan). This is triggered by the pipeline or by a brainstorm handoff. It is the reference for how to author planning documents in the rad-orc workflow.

Use this still if you are a coder agent and have this skill assigned. It is the reference for how to perform a coding task.

Use this still if you are a coder agent and have this skill assigned. It is the reference for how to perform a coding task.

Use this skill to help the user generate beautiful brainstorm/summary visuals, UI mockups and wireframes, and architecture / data-flow / sequence / state diagrams. Trigger when the user wants to *see* something: "make me a wireframe / mockup", "diagram this architecture / flow / sequence", "turn this into a visual or HTML summary", "show me a visual" or "visualize the goals". If you are struggling to align with the user, a visual is often a good way to get on the same page.

Use this still if you are a reviewer agent and have this skill assigned. It is the reference for how to perform a code review at task, phase, or final scope.

Build the Next.js dashboard UI from source and redeploy it to the running install at ~/.radorc/ui for local testing. Packs a standalone bundle via harness-installers/shared/build-helpers/emit-ui-bundle.js, stops the detached UI server, atomically swaps ONLY ~/.radorc/ui (never the full installer hydrate, so config/registry/templates are untouched), and restarts. Use when asked to "build and deploy the UI", "rebuild the dashboard", "redeploy the UI to ~/.radorc/ui", or to see local ui/ source changes in the real detached server.

Use this skill to start the rad orchestration dashboard UI. It is the reference for how to launch the UI and obtain the URL. Use it when the conversation indicates the user wants to start the dashboard. This is commonly used during brainstorming, planning or executing a project or sharing a visual diagram or markdown document with the user. It provides a visual overview of the project and its current state, and allows the user to interact with the pipeline in a more intuitive way.

Brainstorm and refine project goals through collaborative ideation. Use when exploring problem spaces, validating concepts, building consensus on what to build, or early-stage project definition. Trigger when the user talks about brainstorming, goal-setting, idea generation, or early-stage project definition.

Run a project through the rad orc execution pipeline. Use it when after /rad-plan has completed and ready to execute or when the user indicates they want to execute or run a project. It can also be used when resuming a rad-orc project that was previouisly executed but not completed.

Use this when the radorch.mjs pipeline envelope returns ok: false, when an agent produces invalid output, or when manual intervention is needed or any other issue occurs during a pipeline execution.

Use this skill when the user references a project that you're not aware of or need additional context about. It is the reference for how to answer questions about project identity, relationships, project worktree location and how projects interrelate. You may uncover important context for the current project at hand if you know how it relates to other projects. Also consider using this when a user is creates a new project as it might need to have a relationship created to other projects.

Use this skill whenever a task might involve code beyond the current working directory — when you're figuring out where code lives, scoping work that may span multiple repositories, or about to act as if the current repo is the whole system. Use it whenever the user asks to register, bind, describe, group, or manage repositories and repo-groups.

Use this when you are having issues with the rad orchestration dashboard UI. It is the reference for how to check the status of the UI and troubleshoot issues with it.

Use this skill to stop the rad orchestration dashboard UI. It is the reference for how to stop the UI and report the result. This is useful to invoke if the user is having issues using the UI. It may help to stop the UI and restart it to resolve issues. It is also useful to invoke if the user wants to stop the UI for any reason.

Start the radorch dashboard UI dev server (next dev, live reload) from the source repo, auto-building the @rad-orchestration/* lib dist first.

Stage the greenfield rad-orchestration Claude plugin as a local dogfood marketplace. Builds the plugin via the greenfield build chain, copies the build output into a sibling dogfood marketplace tree, writes an ephemeral marketplace.json, and prints the exact /plugin marketplace add + /plugin install commands for the user to run in a fresh Claude Code session. Use when asked to "dogfood the claude plugin", "test the greenfield plugin install", or "stage the plugin for manual install".

Stage the greenfield rad-orchestration Copilot CLI plugin as a local dogfood marketplace. Builds the plugin via the greenfield build chain, copies the build output into a sibling dogfood marketplace tree, writes an ephemeral marketplace.json at .github/plugin/marketplace.json, and prints the exact `copilot plugin marketplace add` + `copilot plugin install` commands for the user to run in a fresh Copilot CLI session. Use when asked to "dogfood the copilot cli plugin", "test the greenfield copilot plugin install", or "stage the copilot plugin for manual install".

Stage the greenfield rad-orchestration Copilot VS Code plugin as a local dogfood marketplace. Builds the plugin via the greenfield build chain, copies the build output into a sibling dogfood marketplace tree, writes an ephemeral marketplace.json at .github/plugin/marketplace.json, and prints the exact install instructions for the user to run in a fresh VS Code session — either via the `Chat: Install Plugin From Source` Command Palette entry or the `chat.pluginLocations` user-settings form. Use when asked to "dogfood the copilot vscode plugin", "test the greenfield copilot vscode plugin install", or "stage the copilot vscode plugin for manual install".

End-to-end installer test for a contributor: builds a local tarball from installer/, runs the installer via npx for the chosen harness, bootstraps the plugin, and verifies via doctor and sha256 manifest check. Use when asked to "test the installer", "build and install the harness", or "verify the install".

Create new custom agents (.agent.md) for the orchestration system. Use when asked to "create an agent", "make a new agent", "scaffold an agent", or when adding a new specialized role to the orchestration pipeline.

Create new Agent Skills for GitHub Copilot from prompts or by duplicating this template. Use when asked to "create a skill", "make a new skill", "scaffold a skill", or when building specialized AI capabilities with bundled resources. Generates SKILL.md files with proper frontmatter, directory structure, and optional scripts/references/assets folders.

Pretends to be an orchestration-system skill. Must be excluded from the manifest by the rad- prefix filter.

Repo-local linting conventions for the rainbow project. Use when authoring tests or test commands. Mandates `npm run rainbow-lint` as the canonical lint command and the `assertRainbow(...)` helper signature in test files.

Manual scaffolding helper — not for model invocation.

Test runner conventions for the foo package. Use when planning tests inside packages/foo/. Mandates the `foo:vitest` npm script as the test command and a `__foo__` test-file suffix.