design-screenshot-spec
Skill7 starsChanged 21 days agoApache-2.0View on GitHub
Inspect Figma links through MCP or browser computer use, or analyze screenshots and selected-layer properties. Produce frontend specs and clarify unresolved product behavior before implementation.
In this repository23 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-rules-skill-authorSkill
- agent-self-improvementSkill
- execution-plan-managerSkill
- frontend-architecture-plannerSkill
- frontend-bugfix-debuggerSkill
- frontend-design-directorSkill
- frontend-design-intelligenceSkill
- frontend-layout-implementerSkill
- frontend-linter-managerSkill
- frontend-prototype-explorerSkill
- frontend-quality-reviewerSkill
- frontend-refactor-surgeonSkill
- frontend-visual-qaSkill
- goal-plannerSkill
- greenfield-project-builderSkill
- loop-workflow-plannerSkill
- mcp-toolchain-managerSkill
- pattern-library-managerSkill
- project-context-adapterSkill
- project-onboarding-adapterSkill
- webdev-feedbackSkill
- webdev-kit-updaterSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 65
Read project/design-reference-profile.md when present.
- Very specific: line 69
After extraction, read references/product-behavior-review.md before formulating product questions or an implementation handoff.
- Distinctive: line 84
Follow host tool instructions and permissions.
- Distinctive: line 99
Treat each screenshot or design-frame size as a reference coordinate system, not as a production container cap.
- Distinctive: line 114
Prefer exact properties only for the same node, state, mode, and revision.
- Distinctive: line 149
Keep detailed per-component evidence in the spec.
- Distinctive: line 154
Every concrete value must cite its source, selection/state, and confidence.
≈ “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
Design Screenshot Spec
Purpose
Inspect live Figma designs or supplied screenshots, property panels, assets,
and notes to produce a traceable Design Implementation Spec. Review the
observed design as a product flow and resolve unspecified design and behavior
decisions with the user before handing affected work to an implementer.
The existing skill name remains stable for bundle compatibility.
When To Use
- The user sends screenshots of Figma frames, components, screens, or inspect panels.
- The user asks to analyze a visual design before implementation.
- The user provides a Figma file, frame, component, or prototype link and asks to inspect it or implement its design.
- The next step is frontend layout work and the implementer needs a structured spec.
When Not To Use
- The user asks to edit the design canvas, create files or whiteboards, generate a design system, or write Code Connect mappings.
- The user already provides a complete
Design Implementation Specand asks to implement it. Usefrontend-layout-implementer.
Required Context
- Read
AGENTS.md. - Confirm the classified task is
design-specor design intake for later frontend implementation. - Read
common/approved-patterns.md. - Read
common/anti-patterns.md. - Read
project/design-reference-profile.mdwhen present. - Read
references/spec-extraction-checklist.md. - For live links, read
references/design-source-inspection.md. - For effects or a visual-effect mismatch, read
references/effect-evidence.md. - After extraction, read
references/product-behavior-review.mdbefore formulating product questions or an implementation handoff.
Tool Contract
- Resolve
live_design_sourceanddesign_reference_filesthrough the current tool registry andtool-capabilities-manifest.jsonfor the active source. - A supplied link triggers scoped read-only inspection automatically: use a
permitted provider selected by the current explicit role/tool instruction;
otherwise use callable Figma MCP first with browser/computer-use fallback.
Apply
common/webdev-role-selection.mdfor an explicitly selected role. - Browser fallback must support screenshots and pointer interaction with the canvas and property panel. Search results or an HTTP fetch are not inspection.
- With screenshots and no link, inspect those images directly; do not require a Figma connection. Never invent a tool, session, node, or successful read.
- Follow host tool instructions and permissions. Do not install tools, change access settings, edit the source canvas, or write mappings during intake.
Workflow
- Inventory links, screenshots, inspect panels, exports, notes, target screens, and existing user decisions. Identify the requested frame and state scope.
- Inspect live links using
references/design-source-inspection.md; use supplied screenshots directly when no link exists. If neither source can be inspected, request the smallest missing reference and report the limitation. - Group evidence by screen, component, variant, state, and viewport. Record source ID, node or selected-layer identity, frame size, image crop/zoom when known, access path, and inspection coverage. Match property panels to their selected elements before assigning values. Inspect every in-scope component; reuse proven identical instances while recording overrides and state changes.
- Treat each screenshot or design-frame size as a reference coordinate system, not as a production container cap. Record whether containment is explicitly visible and capture desktop edge anchors plus expected behavior beyond the reference viewport.
- Extract visible layout, hierarchy, typography, color, spacing, sizing, radius, shadow, assets, states, and responsive behavior.
- For typography, prefer copied inspect values or selected-text properties for
font-family,font-size,font-weight,line-height, color, alignment, max width, and wrapping. When properties are unavailable, estimate from the screenshot and mark each estimate asscreenshot-inferred. - For spacing, separate outside margins, section rhythm, inter-component gaps, container padding, and internal control padding instead of merging them into one generic gap value.
- Mark values
source-provided,screenshot-inferred, orunknown, with the exact source/selection and units; keep user decisions separately traceable. - Prefer exact properties only for the same node, state, mode, and revision. Reinspect mismatches; ask the user which reference governs unresolved conflicts.
- Apply
references/product-behavior-review.md: map observed actions and states, inspect available prototype transitions, then ask about every unresolved product/design decision needed for the requested scope. Keep recommendations unaccepted until the user answers; do not invent defaults. - Produce a draft spec with a decision register. Mark affected slices blocked until answers settle their dependencies. Continue independent inspection; never treat silence or a question limit as approval.
- Hand off the resolved scope only when implementation was requested. Use
frontend-design-directorfor requested redesign or visual choices that need proposals; source-fidelity work does not itself authorize redesign.
Output Contract
Final response: return only facts that affect the user's understanding, confidence, or next action. Omit empty fields and workflow narration.
Return a Design Implementation Spec with these sections:
Source InventoryComponent Evidence And CoverageScreen And Component ScopeLayout StructureTypographyColor And Effects(owning layers, ordered effects, bindings/modes, evidence states, proposed translation and unresolved differences)Spacing And SizingAssetsStates And InteractionsProduct Questions And DecisionsResponsive BehaviorAccessibility NotesImplementation Acceptance CriteriaConfidence And Unknowns
Keep detailed per-component evidence in the spec; user-facing questions should name the relevant component, observation, missing decision, and consequence.
Validation Gates
- Every concrete value must cite its source, selection/state, and confidence.
- A link-only request must attempt an available read path before requesting screenshots. Missing MCP alone must not block an available browser fallback.
- Each in-scope component/state is inspected or explicitly marked missing, unreadable, or blocked; an overview image is not complete extraction.
- Mixed selections and parent properties must not be attributed to a child.
Source Inventorymust record each screenshot's viewport or frame size when visible or provided.Spacing And Sizingmust distinguish measured or estimated margins, section rhythm, inter-component gaps, container padding, and internal control padding with confidence labels.Typographymust distinguish inspect-provided text properties from inferred font family, size, weight, line height, color, alignment, and wrapping.Responsive Behaviormust include viewport-aware notes based on supplied screenshot widths, desktop edge anchors, and behavior beyond each reference viewport, and must mark inferred intermediate behavior explicitly.- Screenshot or design-frame dimensions must not become runtime
max-width, fixed container dimensions, or equivalent caps without source evidence of a centered container and an explicit responsive rationale. - Missing states, assets, breakpoints, and token names must be explicit. Missing effect fields are unknown; inspect ancestors, instance overrides and actual effect types before proposing CSS. Never replace unread values with defaults.
- Unresolved product behavior, visual deviations, and inferred responsive or motion choices must be asked back and block their dependent implementation.
- A static screenshot does not establish transitions, timings, hidden forms, validation, persistence, or error recovery.
- The spec must be usable by
frontend-layout-implementerwithout guessing the target layout intent. - Report actual MCP/browser/image evidence and access gaps honestly; source inspection alone does not prove the implemented UI works.
Trigger Evals
Should trigger:
- "Here are Figma screenshots with dimensions and colors; write the layout spec."
- "Read these frame screenshots and inspect panels before coding."
- "Create a frontend implementation spec from these design screenshots."
- "Open this Figma node and inspect its components before coding."
- "MCP is unavailable; inspect the linked design in the browser."
- "Which layer do these property-panel values belong to?"
Should not trigger:
- "Implement this existing Design Implementation Spec."
- "Create a new Figma whiteboard."
Reference Map
references/effect-evidence.md- layered effects and implementation evidence.common/webdev-role-selection.md- explicit conversational role preferences.references/spec-extraction-checklist.mdreferences/design-source-inspection.md- live MCP and browser acquisition.references/product-behavior-review.md- evidence-led product questions and decision gates after extraction.skills/frontend-design-director/SKILL.md
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.

