Workflow router

Skill owebboy/maestro/skills/workflow-router

Use when the user asks which Maestro workflow fits a task, or invokes the router directly to choose one based on scope, complexity, and available tools.From its SKILL.md

Install

npx -y skills add owebboy/maestro --skill workflow-router

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.9 KB, as published. Nobody here has run it

Workflow Router

This project uses up to three workflow engines at different scope levels. Pick the right one based on the task. Use slash commands in Claude Code and $skill-name invocations in Codex.

Decision Matrix

Task Signal	Route To	Why
Quick fix, single file, obvious change	Plan Mode (built-in)	Think before coding, no ceremony needed
"Add feature", needs design, TDD, multi-file changes	tracked work item `/track-new`	Creates spec, invokes Superpowers brainstorming → planning → execution
Multi-phase work, needs spec + phased plan, project-level tracking	tracked work item `/track-new`	Tracked weight with spec.md, plan.md, phase checkpoints
Work item in-progress	Implement `/implement`	Continue where the last session left off
`.maestro/inbox.md` has unprocessed bullets	Triage `/triage`	Process raw findings into structured work items
Triaged work item needs codebase context	Review `/item-review`	3 parallel agents enrich the item
Small reviewed item: ≤3 files, testable criteria, no design decisions	Direct `/implement <item-ref>`	Implements the item directly with TDD; on completion sets a terminal status (done) which archives it
Reviewed item needing design or multi-phase work	Advance `/item-advance`	Promote to tracked weight with spec + plan
Need a full codebase health check	Codebase review `/codebase-review`	6+6 parallel agent review/audit pattern
Completed work items need acceptance testing	UAT `/uat-create` then `/uat-run`	Generate and run UAT checklist
Session ending, user is done for now	Wrap-up `/session-wrap-up`	Quality review, item capture, context updates, commit
No `.maestro/config.json` present	Setup `/setup`	Initialize project context (product, tech stack, workflow)
Need project overview	Status `/status`	Work item summary, current focus, next actions
Work item housekeeping needed	Manage `/manage`	Archive, restore, delete, rename, cleanup
Need Codex/AGENTS.md interop, or CLAUDE.md changed and AGENTS.md is stale	AGENTS.md sync `/agents-md-sync`	Generate/refresh AGENTS.md from CLAUDE.md for Codex

Availability Detection

Before routing, check what's installed:

Plan Mode: Always available in Claude Code. In Codex, use a brief inline plan for small work or $track-new for tracked planning.
Superpowers: Detect using the detection procedure, checking both plugin-prefixed and bare forms. If found, /track-new and /implement use it as the execution engine. If not, they fall back to inline brainstorming and TDD.
Maestro: Check if .maestro/config.json exists. If not, suggest /setup for projects that would benefit from Maestro's work-item pipeline. If present, read the adapter field and load the capability flags from .maestro/adapters/<adapter>.md — the unified work-item pipeline (triage → review → [advance] → implement) is available regardless of adapter.
Hooks: Some workflows benefit from hook-driven automation (e.g., SessionStart for context injection). Both harnesses support lifecycle hooks using the same nested schema; Codex hooks are on by default (disable with [features] hooks = false). Codex covers 5 events (SessionStart, PreToolUse, PostToolUse, UserPromptSubmit, Stop); Claude adds SessionEnd, SubagentStop, PreCompact, and Notification. Plan accordingly.

When Multiple Skills Compose

Common compositions:

Triage → review → implement (light item): /triage → /item-review → /implement <item-ref>
Triage → review → advance → implement (tracked item): /triage → /item-review → /item-advance → /implement
Quick fix found during work: Use Plan Mode for the fix, capture as inbox item if out of scope
Review → items → tracked: /codebase-review → /triage → /item-advance all
End of day: /session-wrap-up → quality review, item capture, context updates, commit

When to Use Plan Mode vs light vs tracked

Plan Mode — quick, obvious task. One file, one fix, think-before-coding. No ceremony.
Light work item — any non-trivial task that needs triage, enrichment, and a single /implement pass. Goes through triage → review → implement without a full spec/plan.
Tracked work item — anything that benefits from a spec, a design phase, or tracked progress. The /item-advance or /track-new flow writes spec.md + plan.md and uses Superpowers brainstorming and planning under the hood, so you get structured design + TDD + subagent orchestration automatically. Use whenever the work spans multiple files/sessions or needs a design decision.