Sb Figma

sb-figma — Figma → production Storybook delivery

sb-figma owns the DELIVER stage of the Figma→Storybook lifecycle: an approved design becomes production Storybook. It is Figma-aware (uses the Figma MCP), but it is not the exploration skill.

> The lifecycle line (load this first). Undecided / trying options → sb-explore (Lab sandbox, > iterate against a Figma node) → sb-ship (graduate the Lab experiment). Already approved in Figma → > sb-figma (deliver direct to prod). Both touch Figma MCP; the difference is exploration vs delivery. > If the user is still deciding, hand off to sb-explore. See references/figma-token-sync.md.

Job 0 — capture the MCP output (always, before anything else)

> Force the NATIVE structured tools — never work from a screenshot. A screenshot is pixels; it cannot > give you variables, styles, or component props. The full picture comes ONLY from the native MCP tools, and > you must pull all three categories before building or connecting: > - Variables → get_variable_defs (the token values: color/spacing/type/effect, resolved). > - Components + styles → get_design_context (the reference code, applied styles, props/variants). > - Structure → get_metadata (the node tree; on truncation, drill to child node-ids — never give up at > the parent). > get_screenshot is visual reference only — for an eyeball diff after you've built from the structured > data. NEVER read tokens, props, or layout off a screenshot. If a tool returns "nothing selected" on a page > id, drill to a concrete component node (a page is not a layer).

Scripts can't call the Figma MCP, and an MCP result lives only in the agent's context — ephemeral, gone on the next session, absent headless. So every Figma MCP call you make, persist it through the universal store before using it. This is what makes the pipeline reproducible and iterable.

# any JSON-returning tool — pipe its output straight in:
  | node scripts/capture-figma.mjs --tool get_variable_defs   --file  --node  --from-mcp -
 | node scripts/capture-figma.mjs --tool get_design_context  --file  --node  --from-mcp -
       | node scripts/capture-figma.mjs --tool get_metadata        --file  --node  --from-mcp -
  | node scripts/capture-figma.mjs --tool get_code_connect_map --file  --node  --from-mcp -
# get_screenshot returns an image — save it, then register the file:
node scripts/capture-figma.mjs --tool get_screenshot --file  --node  --image /tmp/frame.png
# see the whole inventory (degrade / iterate):
node scripts/capture-figma.mjs --list

Store layout: .storybook/figma/manifest.json + .storybook/figma//.json (images keep their ext). Re-capturing a (tool,node) overwrites — diff against git to see what moved in Figma. Downstream steps read the store, never re-call MCP.

MCP realities (field-verified against the Detections file, 2026-06-22)

What the live Figma MCP actually returns — the scripts already handle these; know them so you don't fight the output:

• get_variable_defs is a FLAT { "name": "value" } map (not nested DTCG, not an array). Colors come **already

resolved to hex ("semantic/background":"#fbfcfc"), incl. 8-digit alpha ("#e4e5e580"). Numbers are bare strings ("spacing-2":"8", "wght/semibold":"650"). Typography is an opaque Font(family: …, size: …) string** — pull-figma-variables parses it to {family,size,weight,lineHeight,…} (a field may itself be a var-name ref like size/text-lg). Shadows are Effect(…) → the effect family. classify() buckets by VALUE.

• get_metadata returns XML and TRUNCATES on large frames (a table view blew past the token limit). When it

truncates: read the child node ids from the partial XML and get_design_context/get_metadata the sub-node, not the parent. Capture each sub-node to the store so you never re-fetch.

• Node-ids: the URL uses 1-6965, the API uses 1:6965. capture-figma canonicalizes to dash form, so a

capture is found whichever way it's passed next. Always pass the node from the URL as-is.

• Persistence + screenshot↔node linking are solved by Job 0 — the store keeps every output with its node-id in

the manifest, so you don't hand-cross-reference screenshots or re-call MCP (both were real friction before).

Two jobs (both consume the Job-0 store)

Job 1 — foundation tokens → Foundations/Colors|Tokens|Type (sb-figma writes these directly)

No other skill maps Figma variables to code tokens, so sb-figma owns the foundation parity end to end.

1. Normalize the captured variables — after Job 0 stored get_variable_defs, point the normalizer at the

stored file: node scripts/pull-figma-variables.mjs --from-mcp .storybook/figma/get_variable_defs/.json --out .storybook/figma-variables.json (with no --from-mcp, it reuses the last --out cache — the headless degrade path). Captures colors, spacing, and type (wght/text/leading).

1. Build parity — node scripts/build-token-parity.mjs --variables .storybook/figma-variables.json --css --out .storybook/figma-token-parity.json.

Resolves OKLCH channel triplets → hex, maps each Figma semantic/* → the project's --token (following var() alias chains), and emits a { token: { figmaVar, figmaHex, mapsTo, drift } } map for all three families. Pass figmaParity to TokenMatrix (it reads figma-token-parity.json itself) to surface drift right in the color table's issue column — figma Δ with code #X vs figma #Y on hover — so design↔code parity lives next to the token, not only in docs/figma-token-parity.md.

1. Wire the foundation stories — Colors.stories.tsx / Tokens.stories.tsx (+ a Type group) read

figma-token-parity.json and pass figmaVar/figmaHex into TokenMatrix rows (and the spacing/type sections). The fields are optional — a project with no Figma file renders exactly as before.

1. Drift — the parity map flags Figma-value ≠ code-computed-value (OKLCH→hex tolerance for color, exact

for spacing/type) and lists app-only roles (--popover, --ring, …) as expected, not failures. Report in the sb-health shape.

Job 2 — approved Figma component → production (sb-figma delivers; sb-stories authors)

1. Extract (via Job 0 store) — capture get_design_context for the node (fall back to get_metadata then

a sub-node fetch if truncated) + get_screenshot, then read them back from .storybook/figma/. The variant list + node-id come from the stored design-context, so re-runs don't re-hit MCP.

1. Audit before adding — grep for an existing component covering the same concept; extend it rather than

duplicate (mirrors guardrails §8 step 3).

1. Build the component with approved tokens/primitives only — tokens, not magic numbers; if a Figma

value has no token, stop and ask (it's a missing-token task, route to Job 1 or the user).

1. Author the story by following sb-stories' conventions — materially-different states only (no

Cartesian), a factory when 3+ stories share a shape. Do not reinvent CSF3 rules; load sb-stories.

1. Stamp + embed — node-id in a top-of-file comment, and parameters.design (see Shared plumbing).
2. Validate — light / dark / mobile, screenshot-vs-implementation parity.
3. Size the delivery FIRST — chunk a big board into parts. A multi-artboard feature (a whole flow,

a screen with many sections) blows past the MCP token budget and produces a "too long, what's the status?" mega-pass. Split by artboard / section: deliver + validate + record one part, then the next. The Figma Inventory (step 7) accretes stories across parts (union by id), so an incremental delivery is first-class, not a workaround. Don't attempt the whole board in one turn.

1. Record the delivery in the Figma Inventory — so the stories this feature created don't just scatter

across the taxonomy. Run the recorder (idempotent; re-run per delivery, stories union by id): ``bash "${CLAUDE_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/scripts/record-figma-delivery.py" . \ --figma-url "" [--feature ""] --spec-url "" --node-ids 101-9717 \ --description "" \ --story "Hunts/Hunt Packs:hunts-hunt-packs--default:component" # repeat per story you created ` Then ensure the root surface exists: scaffold once with scaffold-wrapper.sh --figma, drop the figma-inventory.stories.tsx template (title Figma Inventory), add **one export per feature** (export const Hunts = { args: { feature: 'Hunts' } }), and pin it to the top in .storybook/preview via options.storySort.order: ['Figma Inventory', '*']. The FigmaInventory wrapper reads figma-inventory.json and renders the index + each feature's board link + the stories it brought in. (Spec: docs/specs/2026-06-23-figma-feature-inventory.md`.)

Job 3 — Connect: Storybook → Figma (the reverse direction, code→design)

The reverse of Jobs 1–2: push the code's truth back to Figma so Dev Mode shows your real components. Driven from the Storybook side, fed by context the bundle already has — no new scanners.

1. Gather — for each component that has a story AND a Figma node (the parameters.design node-id from

deliverable #8): its used tokens (from figma-token-parity.json), real prop values (from component-usage.json), variants (from the CSF3 stories), and — from get_metadata — its Figma variant properties ({type:[root,nest-1,nest-2,nest-3]}) and modes ([Light,Dark]). Assemble a comps.json (variantProperties + modes + optional propAliases to rename a Figma prop, e.g. type→depth). build-code-connect maps each variant property → a code prop with its value enum, and carries the modes as the theme dimension — so the Code Connect mapping drives a variant-rich component from Figma's variant picker.

1. Build the payload — `build-code-connect.mjs --components comps.json --parity .storybook/figma-token-parity.json

[--usage .storybook/component-usage.json] --out .storybook/code-connect.json. It enriches each token with its figmaVar` + value, canonicalizes node-ids, and emits a reverse-parity report: components with no Figma node, drifted tokens, tokens not in the parity map.

1. Push (the agent, via MCP) — send_code_connect_mappings / add_code_connect_map with the

mappings[] from code-connect.json. This is the primary, low-risk path — it makes Figma Dev Mode show the real code/props/tokens. Gate the push (it writes to Figma): confirm before sending.

1. Generate (opt-in, OFF by default) — for a component in componentsWithoutNode, generate_figma_design

from the code + story (foundation tokens supplying real values). It creates design artifacts, so only on explicit user intent.

The loop closes: the parameters.design node-id added when delivering (Jobs 1–2) is what Job 3 reads to connect back. Capture any get_code_connect_map / suggestions to the Job-0 store like every MCP call.

Shared plumbing — the Figma design embed (Docs)

Every story sb-figma touches gets the design source preserved on its catalog page:

parameters: { design: { type: 'figma', url: 'https://figma.com/file/?node-id=' } }

This is @storybook/addon-designs (the "Design" tab — the same mechanism sb-explore uses for frames). Plus a node-id stamp + link in parameters.docs.description. sb-explore and sb-stories can reuse this snippet whenever the node-id is known.

Boundaries — never duplicate a sibling's verb

• Exploration / "try a v2" / undecided → sb-explore (Lab). sb-figma is for approved designs only.
• Graduating a Lab experiment → sb-ship (preserve cp, rewrite callsites). sb-figma delivers from

Figma, not from a /explore/ experiment.

• Documenting an existing code component (no Figma) → sb-stories. sb-figma calls sb-stories' rules to

author; it never reimplements them.

• Code-internal token health / orphans → sb-health / sb-inventory (Figma-free). sb-figma adds the

design↔code parity those can't see.

Inputs the agent must resolve first

• Figma file id + variables node id (the variables/styles spec frame) — ask if not pasted.
• Token CSS path(s) — where --token: declarations live (e.g. src/styles/**/*.css).
• Is the design approved? If the user is still iterating → stop, route to sb-explore.

Next

Run Job 1 (tokens) first so components built in Job 2 consume real, parity-checked tokens. Append progress to .storybook/audit/status.md for clean resume. Full call sequence + the OKLCH→hex notes: references/figma-token-sync.md.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: strongeron
• Source: strongeron/storybook-workbench
• License: MIT
• Homepage: https://github.com/strongeron/storybook-workbench

Install and usage instructions live in the source repository linked above.