Skip to content
UXClaim
Design Ops

Goal Workflow Designer

Interrogates you until your /goal prompt is precise enough to work, using a five-element framework and rubric SOP.

What it does

Goal Workflow Designer is a Claude Code skill that coaches you through writing precise, executable specs for Claude’s /goal feature and Dynamic Workflows. Rather than running tasks itself, it interrogates vague requirements and shapes them into detailed prompts that keep agents iterating productively instead of quitting early.

How it works

The skill operates in two modes:

Depth (/goal): Refines a single task through five elements (outcome, verification, constraint, iteration policy, error handling), plus a six-step rubric for subjective work. Outputs a goal prompt ready to paste.

Breadth (workflow-shaper): Evaluates many-unit tasks using a WORTH-IT gate (five criteria), then recommends fan-out shapes (parallel, pipeline, map-reduce) and emits a workflow spec—or refuses if the task isn’t suited to workflows.

Use cases

  • Design reviews: Define measurable criteria before asking agents to critique work
  • Content audits: Specify checks to run across dozens of pages or components
  • Iterative refinement: Lock down what “done” means so agents don’t shortcut to wrap-up
  • Cross-functional handoffs: Package vague briefs into precise specs for teams or agents

Who benefits

Product designers, UX researchers, and design leaders who use Claude Code to automate checks, audits, or iterative work. Especially valuable when context window anxiety causes early quitting or when coordinating multi-agent tasks.

Frequently asked questions

What's the difference between /goal and workflow-shaper?
/goal shapes one task deeply into a precise spec that keeps one agent iterating. workflow-shaper shapes many-unit tasks into parallel or pipeline specs. Count units: one thing to perfect → /goal. Same check across many things → workflow-shaper.
How do I install goal-workflow-designer?
Paste this into Claude Code: 'Read https://raw.githubusercontent.com/dragon375014/goal-workflow-designer/main/INSTALL.md, then install both skills into ~/.claude/skills/.' The agent auto-installs goal and workflow-shaper. Or manually: git clone the repo, copy skills/goal/ and skills/workflow-shaper/ to ~/.claude/skills/, and add triggers to ~/.claude/CLAUDE.md.
Can I use /goal without answering all questions?
The skill is intentionally strict: it interrogates vague answers and won't let you skip elements. If you give 'make it better,' it asks for a measurable standard. This friction is the point—it prevents underspecified goals that cause agents to quit early.
What does the five-element framework cover?
Outcome (what success looks like), verification (how to check it), constraint (limits or guardrails), iteration policy (how many rounds, when to stop), and error handling (what to do if something breaks). For subjective work, it also builds a six-step rubric SOP.
When should I use workflow-shaper instead of /goal?
Use workflow-shaper when you need the same check or change across many independent units (e.g., audit 50 pages, review 12 designs, fix typos in 100 docs). The skill tests eligibility via a WORTH-IT gate and refuses if the task is better suited to /goal or a plain skill.
Does goal-workflow-designer run my task or just design the spec?
It only designs. The skill is a front-end coach: it interrogates you, outputs a precise prompt or workflow spec, and you paste that into /goal or Dynamic Workflows. The skill never executes the actual work.
What is the handoff pattern?
A mechanism for resuming interrupted Q&A sessions. Use /goal --resume <draft-name> to load a saved partial goal, skip already-answered elements, and keep iterating. Useful when context window fills mid-session or you need to step away.
Can I use this skill for non-design tasks?
Yes. Goal Workflow Designer applies to any task that benefits from precise specs: legal review, copywriting, data analysis, code refactoring, audits. The five-element framework and rubric SOP are task-agnostic. It's built for product/design teams but works wherever you need rigorous spec design.

Glossary

WORTH-IT gate
A five-criteria filter in workflow-shaper that evaluates whether a many-unit task is eligible for a Dynamic Workflow. Returns green (go), yellow (middle path), or red (not workflow-suited). Prevents wasting tokens on tasks better handled by /goal or plain skills.
/goal
A Claude Code feature that loops a single agent iteratively toward a target. Works best with precise specs. goal-workflow-designer shapes vague tasks into those precise specs.
Five-element framework
The core structure for /goal specs: outcome (what done looks like), verification (how to measure it), constraint (guardrails), iteration policy (rounds and stopping rules), error handling (failure recovery). Ensures agents know when to stop and what to optimize for.
Rubric SOP
A six-step standard operating procedure for subjective tasks (design critique, content audit, quality judgment). Defines measurable criteria so agents can grade their own work and iterate, rather than inventing their own definition of 'good.'
Context window anxiety
LLM behavior when context fills: agents shortcut to wrap-up and stop iterating, even mid-task. Precise specs and rubrics reduce this by giving agents clear stopping rules and measurable targets, so they don't premature-exit.

More in Design Ops

All →
Design Ops

AI Code Refactoring

Disciplined 5-phase workflow for refactoring AI-generated code with safety nets, design systems, and tests.

chake111
Design Ops

Backend Design

Auto-scaffold a matching backend from your frontend by analyzing screens, components, forms, and API calls.

jwolfsohn
Design Ops

Claude Code Design AI

AI-powered tool that converts design screenshots and wireframes into production-ready React components with Tailwind CSS styling.

Corn7012
Design Ops

Claude Codex Handoff

Three-phase workflow (plan→implement→review) pairing Claude Code specs with Codex implementation and independent adversarial review for complex code changes.