gabimoncha/oh-my-t3/omt-design

Claude Code and agent skill by gabimoncha/oh-my-t3

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Design pass in T3 Code. Use when the user asks for designs, mockups, screens, UX, or a design pass, or when an omt-plan or omt-ulw run has a requested design pass before its UI units. A main thread on Claude Opus draws the profile's variant count in a new Paper file itself; any other model sends the drawing to one routed design child. Shows the screenshots by screen, lets the owner pick one variant per screen and steer until they approve, writes a design spec that UI units cite, and can build t…

Install

npx vskill@latest install gabimoncha/oh-my-t3/omt-design

SKILL.md

Design

You get designs drawn, apply the owner's words until the owner approves, and write the design spec. The owner is the design review. No model reviews the design files or the screenshots.

Steps

0. Check your mode

Call t3_thread_configure with no arguments (delegation.md). A design pass writes files, so in plan mode ask the user to switch this thread to default mode.

Pick the drawer by roles.md: run omt route design --json. When its delegateTask.target.model is your model, you draw. Otherwise one routed design child draws.

Done when: you run in default mode and you know the drawer, or the user has the question.

1. Frame and slug

Write down the goal, the users, the platform, the screens and flows, the constraints, and the existing brand, from the code and assets. Ask the owner only what you cannot find, in one batch, with options and your recommendation.

Pick a slug that matches ^[a-z0-9][a-z0-9-]{0,63}$. The design files go under docs/design/<slug>/, and the spec is docs/design/<slug>.md. The ledger plan is the plan slug when omt-plan or omt-ulw called this skill, else design-<slug>.

Done when: the frame and the slug are in your reply.

2. Draw the variants

When this skill runs alone, run omt git start --run <slug> before the first file. Step 7 builds under the plan <slug>, so the same start record covers the design files.

Draw workflow.design.variants distinct variants of each screen, or the count that the owner asked for. Each variant has its empty, loading, and error states. Use the tool that workflow.design.tool names:

  • paper: a new Paper file named after the project, when Paper's tools load. Reuse no open file. Without Paper's tools, use HTML mockups.
  • html: self-contained HTML mockups.

Export a PNG of each frame to docs/design/<slug>/, by roles/design.md. When Anthropic's optional frontend-design skill is installed, load it, and put its path in a child's brief.

A child draws as unit DS<k>, dispatched by Dispatching a child with --scope docs/design/<slug>/ and mode: "wait". Its task holds the frame, the variant count, and the tool.

Done when: every variant of each screen has a PNG under docs/design/<slug>/, or the owner knows why no drawer can run.

3. Owner checkpoint

Show every screenshot as a Markdown image with its absolute path, such as ![Home, variant A](/abs/path/docs/design/<slug>/home-a.png), grouped by screen, with one line on each variant. Ask in one batch which variant to keep for each screen and what to change. End the turn.

Done when: the owner has the screenshots and the question.

4. Steer

Apply the owner's words verbatim to the chosen variants, in the same Paper file or mockups. When you draw, change the frames and export again. When a child draws, dispatch a new round of the same unit with the owner's words in the task and every earlier round in --prior-file. Run one design child at a time. Then go back to step 3.

Done when: the owner approved one variant for each screen.

5. Write the spec

Write docs/design/<slug>.md yourself, also when a child drew. Read exact values from the Paper file or mockup source, not from the screenshots. Use the spec sections of roles/design.md. Name the approved variant of each screen, and quote the owner's last words.

Done when: the spec exists, each screenshot path in it exists, and the owner has its path.

6. Hand off

Each unit that builds this UI cites docs/design/<slug>.md and uses the ui role. Its qa task gets no docs/design/ paths and no visual-match checks: QA checks behavior, and the owner compares the result with the design.

Return the spec path to omt-plan or omt-ulw when one called this skill. Otherwise give the owner the spec path, and offer to build the screens here.

Done when: the caller, or the owner, has the spec path.

7. Build in the same thread

Only when the owner asks, write a LIGHT plan with omt plan new <slug> --light, with one ui unit per screen that cites the spec. Then run omt-ulw steps 7 to 12.

Done when: the omt-ulw step 12 report is with the owner.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
1
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/gabimoncha/oh-my-t3/omt-design/badge)](https://verified-skill.com/skills/gabimoncha/oh-my-t3/omt-design)

Something wrong with this skill or its page? Tell us.