screens
Skill187 starsChanged 5 months agoMITView on GitHub
Use to produce the canonical screen manifest for a UI feature — every screen in every state (default / loading / empty / error / success / validation), with reuse-first component picks — written to docs/features/{slug}/screens.md between api and tasks. Draws per the design-system tool: Figma MCP into the canon file, Pencil MCP into screens.pen, or inline markdown wireframes (code mode — also the degradation when an MCP is unavailable). Triggers on "screens for {slug}", "screen states for {slug}", "draw the screens", "mockups for {slug}", "/sdd:screens {slug}", "екрани для {slug}", "стани екранів {slug}", "намалюй екрани". States derive from §5 ACs + sad.md §6 branches + contract error responses; tasks/implement/review read ONLY the manifest. Hard-refuse if sad.md is missing; skipped when target_surfaces declares no UI surface.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- raaaf/claude-skillstheir own version, first seen Oct 11, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- raaaf/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
- apiSkill
- clarifySkill
- classify-sizeSkill
- configSkill
- data-modelSkill
- decide-adrSkill
- designSkill
- design-systemSkill
- fixSkill
- glossarySkill
- implementSkill
- interviewSkill
- plan-testsSkill
- reviewSkill
- roadmapSkill
- scaffoldSkill
- sequencesSkill
- shipSkill
- specifySkill
- surveySkill
- tasksSkill
- ux-flowsSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 115
Inventing states with no source: Every non-default state traces to an AC / branch / contract error.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: screens
Turns the approved architecture into the screen manifest — the one artifact tasks,
implement and review read for what every screen shows in every state. It runs between api
and tasks: by then the states are derivable, not invented — default plus every state the
spec §5 ACs, the sad.md §6 alt/else branches, and the contract error responses imply
(loading / empty / error / success / validation). Per state it names the components to reuse
from the docs/design-system.md inventory — a NEW: component only with a
why-no-primitive-fits justification. The visuals are drawn per the canon's tool; whatever the
tool, the manifest is the contract — downstream never reads the raw Figma/.pen file.
Optional by surface: its N/A condition (no UI surface in sad.md target_surfaces) lives in
../_shared/size-matrix.md and is evaluated by api's handoff
(carried forward when api itself is N/A). Invoked directly, it always runs.
Question phrasing → ../_shared/ask-style.md. Manifest prose follows
artifact_language — state tokens (default/loading/…), SCR ids, component names and
source-refs stay English → ../_shared/artifact-language.md.
Owner
Designer + frontend lead. The PM confirms the states match the ACs; review later checks the
built screens against this manifest.
Inputs
<slug>— feature slug.- Gate (hard-refuse if missing):
docs/features/<slug>/sad.md—target_surfaces+ the §6 branches feed the state derivation. Absent → STOP: «rundesign <slug>first». - (Expected)
docs/features/<slug>/ux-flows.md— the SCR inventory this manifest details. Absent → soft: offer/sdd:ux-flows <slug>first, or derive the inventory from spec §4 + the SAD with a noted gap in the manifest — never a silent invention. - (Expected)
docs/design-system.md— thetool+ the component inventory. Absent →codemode- recommend
/sdd:design-systemin the handoff.
- recommend
- (Read)
docs/features/<slug>/contracts/(error responses → error states),spec.md§5 (ACs → states),.size/.route(depth + handoff resolution; defaults stated loudly when missing).
Protocol
- Gate + read.
test -f docs/features/<slug>/sad.md→ missing = refuse with the pointer above. Readtarget_surfaces(no UI surface declared → say so and STOP — this stage is for UI features; the upstream handoff normally skips it). Readux-flows.md(or run the soft fallback above),docs/design-system.md(tool+ inventory),sad.md§6,contracts/, spec §5. Readinterview_depth(else medium;--depth=wins) — it governs the per-screen confirm volume. - Derive states per screen. For each
SCR-NNfrom the inventory, list the full state set:default+ every state the ACs / §6 branches / contract errors imply; a state class that genuinely doesn't apply gets one explicitN/A: <reason>row. Pick the components per state — from the design-system inventory by name; aNEW: <name>only when no existing primitive fits, with the one-line why. Confirm per screen Socratically (medium/hard: oneAskUserQuestionper screen — Accept / Fix / Save-as-OQ / Drop; easy: derive + ledger, ask only where the AC↔state mapping is genuinely ambiguous). - Draw per the canon's
tool.figma→ draw into the canon'sfigma_filevia the Figma MCP, record the node-id per state;pencil→batch_designintodocs/features/<slug>/screens.pen, record node ids;code→ inline markdown wireframes in the manifest. Degrade, don't block: the tool's MCP unavailable in this session → fall back tocodemode and name the degradation in the manifest §Source + the handoff (→../_shared/tool-adapters.md). - Write the manifest per screen. At the first screen, bootstrap
./templates/screens.md→docs/features/<slug>/screens.md(frontmattertoolcopied from the canon; §Source). Steps 2–3 run screen by screen; write each### SCR-NNsection (and any §New components rows it adds) as soon as that screen is confirmed and drawn — progressive writes per../_shared/write-and-commit.md. Stampupdated_atafter the last screen. No commit yet — it follows the self-check. - Structural self-check — per
../_shared/self-check.md: re-read the manifest from disk and verify 4 items: (1) everySCR-NNfromux-flows.mdhas a section (inventory fully covered — or the noted-gap fallback is stated); (2) every screen shows ≥default+ (errororempty) — or carries the explicitN/A: <reason>row; (3) every component name exists in the design-system inventory or sits in §New components with a justification; (4) forfigma/pencil, every state row carries a source-ref (node-id) — forcode, the wireframes are present. Fix + re-check ≤2 cycles; surface anything unresolved. - Commit + handoff. On a passing self-check, commit per
../_shared/write-and-commit.md— one commit forscreens.md(screens: <slug> screens.md) and, onpencil, one forscreens.pen(screens: <slug> screens.pen) only whengit statusshows it changed (the Pencil app persists only on the user's Cmd+S — an unsaved.penis named in the handoff, not committed); Figma lives outside the repo, nothing to commit. An unresolved item means no commits. Then emit the stage-handoff block per../_shared/handoff.md— What I did (tool used or the degradation, «self-check: 4/4 pass», the commit SHAs) + Review (docs/features/<slug>/screens.md, +screens.pen/ the Figma file when tool-drawn) + Run next:/clear, then/sdd:tasks <slug>(eachuitask will cite these SCR ids + states).
Definition of Done
docs/features/<slug>/screens.mdexists: §Source per the canon's tool, one section per SCR with the derived state table, components reuse-first (NEW:only justified), §New components filled or explicitly «None».- States are derived — every error/empty state traces to an AC, a §6 branch, or a contract error response; nothing invented, nothing silently missing.
- For
figma/pencil: every state has its node source-ref; forcode: the wireframes are inline. - The manifest is the only thing downstream needs —
tasks/implement/reviewnever open the raw design file.
Anti-patterns
- Happy-path-only screens. A screen with just
defaulthides exactly the states users hit — the error/empty derivation is the point of this stage. - Inventing states with no source. Every non-default state traces to an AC / branch / contract error; a state with no origin is scope creep in a costume.
- Hand-rolling a component the inventory already has — the reuse rule this pipeline exists to
enforce;
NEW:requires the why-no-primitive-fits line. - Blocking on a missing MCP.
codemode is always available — degrade with a named degradation, never refuse the stage over tooling. - Making downstream read the design file. Node-ids are source-refs for humans; the manifest
carries everything
tasks/implement/reviewconsume. - Architecture decisions here. The surface set, SSR-vs-SPA, state management — all
design's; screens details what was declared, it never re-decides.
References & template
./templates/screens.md— the manifest scaffold; inline comments are the generation contract.../_shared/size-matrix.md— the N/A condition (no UI surface) evaluated byapi's handoff.../_shared/tool-adapters.md— the design-tool MCP degradation rule (fall back to code mode, named).../_shared/surfaces.md— the surface declaration this stage gates on.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

