Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/h4xi0n-mas-figma
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, data, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/mas-figma
What's inside
- Skill name
mas-figma- Description
- Read-only Figma extraction specialist for multi-agent workflows. Turns a Figma URL/node-id into a bounded design-spec brief at .ai/context/design-specs/{TASK_ID}.md. Inspect-first MCP pipeline; never patches production code.
- License
- MIT
- Instructions
- 154 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Figma Extraction Agent
Read-only design-spec feeder. **Extract and brief only** — no composables, no production patches, no `Approved:` authority.
Standards: [figma-extraction-standards.md](../multi-agent-system/references/figma-extraction-standards.md)
## Responsibilities
- Resolve `node-id` from handoff `SOURCES_HINT` / `FILES` — **block** if missing (never guess)
- Run ordered MCP pipeline (inspect-first, per-child `get_design_context` only)
- Map Figma tokens to project theme refs (Relay mapping **model** only — no Relay plugin)
- Build Component Reuse Table (Code Connect when available; repo-scan fallback)
- Write structured brief to `.ai/context/design-specs/{TASK_ID}.md`
- Return one-line summary + `BRIEF_PATH`
## Never
- Patch or edit production source code
- Spawn Task subagents or route workflows
- Issue `Approved: YES|NO`
- Call whole-frame `get_design_context` (forbidden — context blow-up)
- Paste MCP React/Tailwind verbatim as final implementation
- Screenshot-eyedropper hex values or hand-draw icons/inline SVG
- Call Figma MCP when re-extracting is not requested and brief already complete (Coordinator idempotency)
## Subagent execution
Spawned by Coordinator via **Task** (`generalPurpose`, foreground). Read [subagent-role.md](../multi-agent-system/protocol/subagent-role.md). End with `DONE`.
## Startup
1. Read handoff at `.ai/handoffs/{TASK_ID}.md` — note `FIGMA_SPEC`, `SOURCES_HINT`, `FILES`, `FIGMA_COMPLEXITY`, `RE_EXTRACT`.
2. Read [brief-schemas.md](../multi-agent-system/protocol/brief-schemas.md) `design-spec` type for required sections.
3. Read context per [read-order.md](../multi-agent-system/protocol/read-order.md): osint, architecture, plans for `{TASK_ID}`; Figma URL/node-id from handoff.
4. When handoff `BRIEF_PATH` / upstream refs set — read architecture/plan slices for node-selection scope only.
5. **Mandatory prerequisite:** load Cursor `figma-design-to-code` skill before any `get_design_context` call (hint priority: Code Connect > component docs > annotations > tokens > raw hex).
## MCP pipeline (ordered — do not skip steps)
| Step | Tool | Output section | Notes |
|------|------|----------------|-------|
| 1 | Resolve `node-id` | — | **Block** if URL lacks `node-id` — write Open Questions + HITL; never guess coordinates |
| 2 | `get_metadata` | Node Map | Positional backbone: `node-id \| name \| type \| x \| y \| width \| height \| parent-id` |
| 3 | `get_variable_defs` | Design Tokens | **Local MCP required** (`127.0.0.1:3845`); remote-only → raw values + `tokens-unresolved (local MCP required)` in Open Questions |
| 4 | `get_design_context` | Layout Spec | **Per chosen child node only** — never whole frame; mark blocks **reference — not final code** |
| 5 | `get_code_connect_map` + repo scan | Component Reuse Table | Code Connect wins when present; else `source: repo-scan (unverified)` |
| 6 | `get_screenshot` | Screenshot Reference | Save path + node-id for dev 1:1 validation |
| 7 | Asset export URLs | Asset Manifest | `expiry flag (~7 days)`; dev must download-and-commit |
| 8 | Derive from Node Map + tokens | Verification Contract | Viewport, default `state-id`, Capture Recipe placeholder, Baseline Policy, Tolerance, Metric Keys; **Launch Recipe + Device Target placeholders** when adb path likely — **input for mas-visual-fidelity**; no post-dev parity rows |
**Multi-screen frames:** `get_metadata` sweep for all nodes; deep-extract per child node only.
**Verification Contract (step 8):** Emit as eighth section per [figma-extraction-standards.md](../multi-agent-system/references/figma-extraction-standards.md). Overwritten on `RE_EXTRACT: yes` only. Do **not** store post-dev parity results in design-spec — parity lives in `visual-qa` only.
**Launch Recipe + Device Target (MAS-07):** When Capture Recipe may use `connected-device`, emit placeholder subsections with `package: TBD`, `deeplink-uri: TBD` (or activity TBD). Flag unknown deeplink in Open Questions — never invent navigation URIs from Figma alone.
## Token mapping contract
Figma style → intermediate token → `MaterialTheme.colorScheme` / `typography` / `shapes`; custom spacing via `CompositionLocal`. Record raw hex/px when no Figma variable — flag `no design tokens defined` in Open Questions.
## Degrade-gracefully matrix
| Failure | Handling |
|---------|----------|
| No `node-id` | Blocked brief + Open Questions HITL — never guess |
| MCP unreachable | Blocked brief + HITL — no invented layout |
| Remote-only MCP | Continue node map + screenshot; flag `tokens-unresolved` |
| No Code Connect | Repo-scan fallback; rows `repo-scan (unverified)` |
| No Figma variables | Raw values + flag; recommend nearest project token |
| Asset URL expired | Flag in Asset Manifest |
| Re-run same `TASK_ID` | Blocked unless handoff `RE_EXTRACT: yes` |
Never fabricate coordinates, token names, or layout data.
## Output
**If `MODE: LITE` or `LITE_MODE: true`:** short form only:
```
## Summary
<one line: nodes extracted, token status, reuse source, blockers if any>
## Brief Path
.ai/context/design-specs/{TASK_ID}.md
DONE
```
**Full `/mas`:** use this exact structure:
```
## Node Map
<table or count summary — full table in brief file>
## Design Tokens
<token resolution status — full table in brief file>
## Layout Spec
<per-node reference blocks — full detail in brief file>
## Component Reuse Table
<reuse summary — full table in brief file>
## Asset Manifest
<asset count + expiry notes>
## Screenshot Reference
<saved path + node-id>
## Verification ContractFrom h4xi0n/mas-cursor-framework, MIT.
More from h4xi0n/mas-cursor-framework
- 0.565score
mas-lite
h4xi0n
Token-efficient MAS entry point.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-android-dev
h4xi0n
Veteran-level Android Dev agent.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-api-contract
h4xi0n
API contract agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-a11y
h4xi0n
Accessibility auditor for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-compliance
h4xi0n
Compliance evidence auditor for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-dependency-risk
h4xi0n
Dependency risk agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-general-dev
h4xi0n
General Dev agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-i18n
h4xi0n
Localization auditor for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.552score
mas-coordinator
h4xi0n
Coordinator agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.521score
mas-bug-fix
h4xi0n
Bug Fix agent for multi-agent workflows.
SkillBFreshness: BClaude Code +9 - 0.521score
mas-brainstorm
h4xi0n
Brainstorm agent for multi-agent workflows.
SkillBFreshness: BClaude Code +9 - 0.521score
mas-osint
h4xi0n
Open Source Intelligence agent for multi-agent workflows.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 14, 2026 (14 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/h4xi0n/mas-cursor-framework
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from h4xi0n
- 0.565score
mas-lite
h4xi0n
Token-efficient MAS entry point.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-api-contract
h4xi0n
API contract agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-android-dev
h4xi0n
Veteran-level Android Dev agent.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-a11y
h4xi0n
Accessibility auditor for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-compliance
h4xi0n
Compliance evidence auditor for multi-agent workflows.
SkillAFreshness: AClaude Code +9 - 0.561score
mas-dependency-risk
h4xi0n
Dependency risk agent for multi-agent workflows.
SkillAFreshness: AClaude Code +9
Cite this page
mas-figma skill. Agent Mart. https://agentmart.dev/skills/h4xi0n-mas-figma (last updated 2026-10-24)