@skilly-hand/skilly-hand 0.26.2 → 0.26.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +30 -0
- package/catalog/README.md +1 -1
- package/catalog/skills/figma-mcp-0to1/SKILL.md +27 -8
- package/catalog/skills/figma-mcp-0to1/agents/canvas-creation-playbook.md +8 -1
- package/catalog/skills/figma-mcp-0to1/agents/install-auth.md +7 -5
- package/catalog/skills/figma-mcp-0to1/agents/tool-function-catalog.md +5 -2
- package/catalog/skills/figma-mcp-0to1/agents/troubleshooting-ops.md +8 -0
- package/catalog/skills/figma-mcp-0to1/assets/client-config-snippets.md +7 -2
- package/catalog/skills/figma-mcp-0to1/assets/prompt-recipes.md +14 -1
- package/catalog/skills/figma-mcp-0to1/manifest.json +3 -3
- package/catalog/skills/figma-mcp-0to1/references/official-tools-matrix.md +25 -5
- package/catalog/skills/frontend-design/SKILL.md +22 -9
- package/catalog/skills/frontend-design/agents/component-designer.md +1 -0
- package/catalog/skills/frontend-design/agents/critique.md +3 -0
- package/catalog/skills/frontend-design/agents/design-context-setter.md +15 -2
- package/catalog/skills/frontend-design/agents/visual-refiner.md +4 -0
- package/catalog/skills/frontend-design/assets/taste-reference-extraction.md +161 -0
- package/catalog/skills/frontend-design/manifest.json +6 -5
- package/package.json +1 -1
- package/packages/catalog/package.json +1 -1
- package/packages/cli/package.json +1 -1
- package/packages/core/package.json +1 -1
- package/packages/detectors/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -16,6 +16,36 @@ All notable changes to this project are documented in this file.
|
|
|
16
16
|
### Removed
|
|
17
17
|
- _None._
|
|
18
18
|
|
|
19
|
+
## [0.26.4] - 2026-05-09
|
|
20
|
+
[View on npm](https://www.npmjs.com/package/@skilly-hand/skilly-hand/v/0.26.4)
|
|
21
|
+
|
|
22
|
+
### Added
|
|
23
|
+
- Added taste-reference extraction guidance to `frontend-design`, with Refero Styles and Impeccable-inspired workflows for turning references into register, visual ingredients, anti-references, and DESIGN.md-ready taste contracts.
|
|
24
|
+
|
|
25
|
+
### Changed
|
|
26
|
+
- Updated `frontend-design` setup, component design, critique, and refinement routing to apply reference-derived taste rules instead of treating visual direction as vague mood language.
|
|
27
|
+
|
|
28
|
+
### Fixed
|
|
29
|
+
- _None._
|
|
30
|
+
|
|
31
|
+
### Removed
|
|
32
|
+
- _None._
|
|
33
|
+
|
|
34
|
+
## [0.26.3] - 2026-05-09
|
|
35
|
+
[View on npm](https://www.npmjs.com/package/@skilly-hand/skilly-hand/v/0.26.3)
|
|
36
|
+
|
|
37
|
+
### Added
|
|
38
|
+
- _None._
|
|
39
|
+
|
|
40
|
+
### Changed
|
|
41
|
+
- Updated the `figma-mcp-0to1` skill for current Figma MCP remote-first setup, supported-client coverage, Figma-provided skills, FigJam write workflows, code-to-canvas support, and access/rate-limit guidance.
|
|
42
|
+
|
|
43
|
+
### Fixed
|
|
44
|
+
- _None._
|
|
45
|
+
|
|
46
|
+
### Removed
|
|
47
|
+
- _None._
|
|
48
|
+
|
|
19
49
|
## [0.26.2] - 2026-05-03
|
|
20
50
|
[View on npm](https://www.npmjs.com/package/@skilly-hand/skilly-hand/v/0.26.2)
|
|
21
51
|
|
package/catalog/README.md
CHANGED
|
@@ -8,7 +8,7 @@ Published portable skills consumed by the `skilly-hand` CLI.
|
|
|
8
8
|
| `agents-root-orchestrator` | Author root AGENTS.md as a Where/What/When orchestrator that routes tasks and skill invocation clearly. | core, workflow, orchestration | all |
|
|
9
9
|
| `angular-guidelines` | Guide Angular code generation, review, and performance tuning using latest stable Angular verification, official Angular skill guidance, and modern framework best practices. Trigger: generating, reviewing, refactoring, or optimizing Angular code artifacts in Angular projects. | angular, frontend, workflow, best-practices | all |
|
|
10
10
|
| `figma-mcp-0to1` | Guide users from Figma MCP installation and authentication through first canvas creation, with function-level tool coverage and operational recovery patterns. | figma, mcp, workflow, design | all |
|
|
11
|
-
| `frontend-design` | Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish. | frontend, design, workflow, ui, motion, greenfield | all |
|
|
11
|
+
| `frontend-design` | Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, taste-reference extraction, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish. | frontend, design, workflow, ui, motion, greenfield | all |
|
|
12
12
|
| `gsap-animation` | Guide GSAP animation implementation using only official GSAP documentation and the official greensock/gsap-skills source material. Trigger: implementing, reviewing, or choosing GSAP for frontend motion, timelines, ScrollTrigger, React useGSAP, JavaScript animation libraries, or advanced UI animation. | frontend, animation, motion, gsap, workflow | all |
|
|
13
13
|
| `motion-animation` | Guide Motion, formerly Framer Motion, animation implementation using only official Motion documentation. Trigger: implementing, reviewing, or choosing Motion for JavaScript animation, React motion components, gestures, scroll animation, layout animation, exit animation, or framework-agnostic UI motion. | frontend, animation, motion, framer-motion, workflow | all |
|
|
14
14
|
| `output-optimizer` | Optimize output token consumption through compact interpreter modes with controlled expansion when complexity, ambiguity, or risk requires more detail. Trigger: minimizing response verbosity while preserving clarity and correctness. | core, workflow, efficiency, communication | all |
|
|
@@ -3,10 +3,10 @@ name: "figma-mcp-0to1"
|
|
|
3
3
|
description: "Guide users from Figma MCP installation and authentication through first canvas creation, with function-level tool coverage and operational recovery patterns."
|
|
4
4
|
skillMetadata:
|
|
5
5
|
author: "skilly-hand"
|
|
6
|
-
last-edit: "2026-05-
|
|
6
|
+
last-edit: "2026-05-09"
|
|
7
7
|
license: "Apache-2.0"
|
|
8
|
-
version: "1.0.
|
|
9
|
-
changelog: "
|
|
8
|
+
version: "1.0.3"
|
|
9
|
+
changelog: "Updated Figma MCP guidance for current remote-first docs, supported-client coverage, Figma-provided skills, FigJam write workflows, code-to-canvas support, and access/rate-limit language; affects setup routing, tool selection, prompt assets, and official matrix"
|
|
10
10
|
auto-invoke: "Installing, configuring, or using Figma MCP from setup through first canvas creation"
|
|
11
11
|
allowed-tools:
|
|
12
12
|
- "Read"
|
|
@@ -66,13 +66,14 @@ Choose subskills by intent:
|
|
|
66
66
|
|
|
67
67
|
## Core Rules
|
|
68
68
|
|
|
69
|
-
- Prefer remote server for broadest feature coverage and
|
|
69
|
+
- Prefer remote server for broadest feature coverage, write workflows, code-to-canvas, and FigJam agent workflows.
|
|
70
70
|
- Treat official Figma MCP docs as the source of truth for official tools, supported clients, permissions, and limits.
|
|
71
71
|
- Keep client-specific helpers separate from official Figma MCP tools.
|
|
72
72
|
- Treat write actions as staged operations, not a single large operation.
|
|
73
73
|
- Use link-based node targeting for reliable design-context extraction.
|
|
74
74
|
- Keep a clear distinction between read context tools and write/canvas tools.
|
|
75
|
-
- For repeated team workflows, reuse prompts and config snippets from `assets
|
|
75
|
+
- For repeated team workflows, reuse prompts and config snippets from `assets/`, and prefer Figma-provided skills when they exist.
|
|
76
|
+
- Use `figma-use` with `use_figma` for Figma Design writes and `figma-use-figjam` for FigJam writes when those skills are available.
|
|
76
77
|
|
|
77
78
|
---
|
|
78
79
|
|
|
@@ -82,9 +83,24 @@ Figma MCP support and skilly-hand installation support are related but not ident
|
|
|
82
83
|
|
|
83
84
|
| Coverage | Agents or Clients | Guidance |
|
|
84
85
|
| --- | --- | --- |
|
|
85
|
-
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot` | Include concrete setup paths
|
|
86
|
-
| Figma-supported but not skilly-hand-native | VS Code, Warp, Augment, Factory, Firebender | Mention as Figma-supported clients, but do not add skilly-hand install assumptions. |
|
|
87
|
-
| skilly-hand-supported but not source-backed in current Figma docs | `
|
|
86
|
+
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot`, `gemini` | Include concrete setup paths when Figma publishes them; otherwise point to client docs/catalog. |
|
|
87
|
+
| Figma-supported but not skilly-hand-native | Amazon Q, Android Studio, VS Code, Warp, Augment, Factory, Firebender, Kiro, OpenHands, Replit | Mention as Figma-supported clients, but do not add skilly-hand install assumptions. |
|
|
88
|
+
| skilly-hand-supported but not source-backed in current Figma docs | `antigravity`, `windsurf`, `trae` | Keep broad `agentSupport`; document that Figma-specific setup may require client documentation or manual MCP config. |
|
|
89
|
+
|
|
90
|
+
## Figma-Provided Skills
|
|
91
|
+
|
|
92
|
+
Prefer Figma-provided skills for workflows they cover:
|
|
93
|
+
|
|
94
|
+
| Skill | Use |
|
|
95
|
+
| --- | --- |
|
|
96
|
+
| `figma-use` | Foundational Figma Design write-to-canvas workflow for frames, components, variables, styles, and auto layout. |
|
|
97
|
+
| `figma-use-figjam` | Foundational FigJam write workflow for boards, stickies, sections, connectors, shapes, tables, and code blocks. |
|
|
98
|
+
| `figma-create-new-file` | Create blank Figma Design or FigJam files before writing. |
|
|
99
|
+
| `figma-implement-design` | Generate production code from a Figma design URL. |
|
|
100
|
+
| `figma-code-connect-components` | Map published Figma components to code implementations. |
|
|
101
|
+
| `figma-create-design-system-rules` | Generate reusable code-generation rules for a project. |
|
|
102
|
+
| `figma-generate-library` | Example workflow for creating or syncing a Figma design-system library from code. |
|
|
103
|
+
| `figma-generate-design` | Example workflow for building screens/views in Figma from code or a design-system-aware brief. |
|
|
88
104
|
|
|
89
105
|
---
|
|
90
106
|
|
|
@@ -117,6 +133,9 @@ claude mcp add --transport http figma-desktop http://127.0.0.1:3845/mcp
|
|
|
117
133
|
# Cursor plugin setup
|
|
118
134
|
/add-plugin figma
|
|
119
135
|
|
|
136
|
+
# Gemini CLI
|
|
137
|
+
# Follow the Figma MCP catalog/client instructions for the Gemini CLI extension.
|
|
138
|
+
|
|
120
139
|
# Verify catalog integrity in this repository
|
|
121
140
|
npm run catalog:check
|
|
122
141
|
```
|
|
@@ -26,7 +26,8 @@ Deliver a first successful Figma canvas result with low risk and clear verificat
|
|
|
26
26
|
- Create one frame.
|
|
27
27
|
- Add one text node.
|
|
28
28
|
- Apply one style/token decision.
|
|
29
|
-
- For
|
|
29
|
+
- For Figma Design, load or invoke `figma-use` when available.
|
|
30
|
+
- For FigJam, load or invoke `figma-use-figjam` when available, then create or update one small board object such as a sticky, section, connector, shape, table, or code block.
|
|
30
31
|
|
|
31
32
|
5. **Validate state**
|
|
32
33
|
- Inspect with `get_metadata` or `get_screenshot`.
|
|
@@ -44,6 +45,12 @@ Deliver a first successful Figma canvas result with low risk and clear verificat
|
|
|
44
45
|
2. Let `generate_diagram` create a new FigJam file, or provide an existing FigJam file key.
|
|
45
46
|
3. Validate diagram structure and labels.
|
|
46
47
|
|
|
48
|
+
### FigJam board-first flow
|
|
49
|
+
|
|
50
|
+
1. Create or choose a FigJam file, using `create_new_file figjam <name>` if a fresh board is needed.
|
|
51
|
+
2. Use `figma-use-figjam` with `use_figma` for small write steps.
|
|
52
|
+
3. Validate with `get_figjam` or `get_screenshot`.
|
|
53
|
+
|
|
47
54
|
### Code-to-canvas flow
|
|
48
55
|
|
|
49
56
|
1. Ask client to start local app capture workflow.
|
|
@@ -8,8 +8,8 @@ Get Figma MCP connected and authenticated, with a verified first tool call.
|
|
|
8
8
|
|
|
9
9
|
| Mode | Endpoint | Recommended | Notes |
|
|
10
10
|
| --- | --- | --- | --- |
|
|
11
|
-
| Remote | `https://mcp.figma.com/mcp` | Yes | Broadest feature coverage, including write-to-canvas, code-to-canvas, diagrams, and file creation. |
|
|
12
|
-
| Desktop | `http://127.0.0.1:3845/mcp` | Only for local selection-based or special org needs | Requires Figma desktop app + Dev Mode desktop MCP toggle; narrower workflow coverage. |
|
|
11
|
+
| Remote | `https://mcp.figma.com/mcp` | Yes | Broadest feature coverage, including write-to-canvas, code-to-canvas, FigJam workflows, diagrams, and file creation. Remote access is documented for all seats/plans, with plan/seat-based limits. |
|
|
12
|
+
| Desktop | `http://127.0.0.1:3845/mcp` | Only for local selection-based or special org/enterprise needs | Requires Figma desktop app + Dev Mode desktop MCP toggle; narrower workflow coverage and paid Dev/Full seat access. |
|
|
13
13
|
|
|
14
14
|
## Codex Setup
|
|
15
15
|
|
|
@@ -37,6 +37,7 @@ Use concrete setup paths for agents that are both Figma-supported and skilly-han
|
|
|
37
37
|
- Cursor plugin: `/add-plugin figma`
|
|
38
38
|
- Cursor manual: add the remote or desktop endpoint in MCP settings.
|
|
39
39
|
- Copilot: use the VS Code MCP route and enable GitHub Copilot in the client.
|
|
40
|
+
- Gemini CLI: follow the Figma MCP catalog/client instructions for the Gemini CLI extension.
|
|
40
41
|
- VS Code: add HTTP MCP server in `mcp.json` using the remote endpoint.
|
|
41
42
|
|
|
42
43
|
See exact snippets in [../assets/client-config-snippets.md](../assets/client-config-snippets.md).
|
|
@@ -45,9 +46,9 @@ See exact snippets in [../assets/client-config-snippets.md](../assets/client-con
|
|
|
45
46
|
|
|
46
47
|
| Coverage | Agents or Clients | Guidance |
|
|
47
48
|
| --- | --- | --- |
|
|
48
|
-
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot` | Use the commands and setup paths above. |
|
|
49
|
-
| Figma-supported but not skilly-hand-native | VS Code, Warp, Augment, Factory, Firebender | Follow Figma/client setup docs; do not assume skilly-hand skill install paths. |
|
|
50
|
-
| skilly-hand-supported but not source-backed in current Figma docs | `
|
|
49
|
+
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot`, `gemini` | Use the commands and setup paths above where documented; otherwise follow the Figma MCP catalog/client docs. |
|
|
50
|
+
| Figma-supported but not skilly-hand-native | Amazon Q, Android Studio, VS Code, Warp, Augment, Factory, Firebender, Kiro, OpenHands, Replit | Follow Figma/client setup docs; do not assume skilly-hand skill install paths. |
|
|
51
|
+
| skilly-hand-supported but not source-backed in current Figma docs | `antigravity`, `windsurf`, `trae` | Keep install support, but verify Figma MCP setup through that client's docs or manual MCP config. |
|
|
51
52
|
|
|
52
53
|
## Verification Checklist
|
|
53
54
|
|
|
@@ -57,6 +58,7 @@ See exact snippets in [../assets/client-config-snippets.md](../assets/client-con
|
|
|
57
58
|
- Desktop-only contexts: `get_metadata` or `get_design_context` on a known node
|
|
58
59
|
3. Confirm the authenticated user has access to the target Figma file.
|
|
59
60
|
4. For write workflows, confirm the authenticated seat can edit the target file.
|
|
61
|
+
5. For Dev seats, confirm the target is a draft or that the workflow does not require edit access outside drafts.
|
|
60
62
|
|
|
61
63
|
## First Prompt After Setup
|
|
62
64
|
|
|
@@ -7,7 +7,7 @@ Pick the smallest correct Figma MCP function for the task, with predictable inpu
|
|
|
7
7
|
## Start Here
|
|
8
8
|
|
|
9
9
|
1. Read [../references/official-tools-matrix.md](../references/official-tools-matrix.md).
|
|
10
|
-
2. Identify whether task is read, write, design-system search, code-connect, diagram, live UI capture, or Make resource context.
|
|
10
|
+
2. Identify whether task is read, write, FigJam, design-system search, code-connect, diagram, live UI capture, or Make resource context.
|
|
11
11
|
3. Prefer read-first calls before write calls on unknown files.
|
|
12
12
|
|
|
13
13
|
## Selection Rules
|
|
@@ -18,9 +18,11 @@ Pick the smallest correct Figma MCP function for the task, with predictable inpu
|
|
|
18
18
|
- Need variables/tokens only: `get_variable_defs`.
|
|
19
19
|
- Need visual reference: `get_screenshot`.
|
|
20
20
|
- Need FigJam extraction: `get_figjam`.
|
|
21
|
+
- Need FigJam creation/editing: `use_figma` with the `figma-use-figjam` skill when available.
|
|
21
22
|
- Need library discovery in Codex before a scoped search: `get_libraries` if the Codex Figma plugin exposes it.
|
|
22
23
|
- Need design-system discovery before creating: `search_design_system`.
|
|
23
|
-
- Need to write/create/update Figma Design
|
|
24
|
+
- Need to write/create/update Figma Design content: `use_figma` (remote write workflow; use `figma-use` skill when available).
|
|
25
|
+
- Need to write/create/update FigJam content: `use_figma` (remote write workflow; use `figma-use-figjam` skill when available).
|
|
24
26
|
- Need first-time live web UI capture: `generate_figma_design`.
|
|
25
27
|
- Need new blank file: `create_new_file` (requires plan/team context such as `planKey` in some clients).
|
|
26
28
|
- Need Mermaid-to-FigJam: `generate_diagram` (creates its own FigJam file unless an existing FigJam file key is provided).
|
|
@@ -35,6 +37,7 @@ Use direct tool-trigger language when selection is ambiguous:
|
|
|
35
37
|
- "Use Figma MCP `search_design_system` before creating any new component."
|
|
36
38
|
- "Use Figma MCP `generate_figma_design` to capture this live web UI to a new Figma file."
|
|
37
39
|
- "Use Figma MCP `create_new_file`, then `use_figma` to add a first frame and typography style."
|
|
40
|
+
- "Use `figma-use-figjam` with `use_figma` to organize this FigJam board into sections and connectors."
|
|
38
41
|
- "Use MCP resources from this Figma Make link and fetch only the files needed for implementation."
|
|
39
42
|
|
|
40
43
|
## Remote/Desktop Caveats
|
|
@@ -33,6 +33,7 @@ Failure observed?
|
|
|
33
33
|
1. Validate Figma URL format and node id.
|
|
34
34
|
2. Run `whoami` to verify authenticated account.
|
|
35
35
|
3. Confirm authenticated user belongs to correct plan and has file access.
|
|
36
|
+
4. For write-to-canvas, confirm the user has edit access; Dev seats are read-only outside drafts.
|
|
36
37
|
|
|
37
38
|
### Rate limiting
|
|
38
39
|
|
|
@@ -41,6 +42,7 @@ Failure observed?
|
|
|
41
42
|
- Dev/Full seats on Professional or Organization plans have higher daily usage.
|
|
42
43
|
- Enterprise Dev/Full seats have the highest documented daily usage.
|
|
43
44
|
- Officially documented exempt tools include `add_code_connect_map`, `generate_figma_design`, and `whoami`.
|
|
45
|
+
- `use_figma` write tools may be documented outside standard read limits, but still require correct seat and edit permissions.
|
|
44
46
|
- Use smaller selections and fewer repeated reads.
|
|
45
47
|
- Prefer `get_metadata` preflight before broad `get_design_context`.
|
|
46
48
|
- Batch intent into fewer, targeted calls.
|
|
@@ -58,6 +60,12 @@ Failure observed?
|
|
|
58
60
|
- Verify partial outcomes using read tools before next step.
|
|
59
61
|
- For `create_new_file`, ensure a plan/team context is available.
|
|
60
62
|
- For `generate_diagram`, do not pre-create a blank FigJam file unless adding to an existing file is the explicit goal.
|
|
63
|
+
- For FigJam writes, use `figma-use-figjam` guidance where available instead of generic Figma Design assumptions.
|
|
64
|
+
|
|
65
|
+
### Expired image URLs
|
|
66
|
+
|
|
67
|
+
- Figma-hosted image URLs returned by MCP can expire.
|
|
68
|
+
- Refresh the design context to get new URLs, or download/save assets into the project and reference local files for durable implementation.
|
|
61
69
|
|
|
62
70
|
## Scoped Retry Pattern
|
|
63
71
|
|
|
@@ -90,6 +90,10 @@ claude mcp list
|
|
|
90
90
|
/add-plugin figma
|
|
91
91
|
```
|
|
92
92
|
|
|
93
|
+
## Gemini CLI
|
|
94
|
+
|
|
95
|
+
Use the Figma MCP catalog/client instructions for the Gemini CLI extension. Figma currently lists Gemini CLI as a supported client for remote and desktop setup.
|
|
96
|
+
|
|
93
97
|
## Cursor MCP config (desktop)
|
|
94
98
|
|
|
95
99
|
```json
|
|
@@ -117,5 +121,6 @@ claude mcp list
|
|
|
117
121
|
## Agent Coverage Notes
|
|
118
122
|
|
|
119
123
|
- Figma-supported and skilly-hand-supported: `codex`, `claude`, `cursor`, `copilot`.
|
|
120
|
-
- Figma-supported
|
|
121
|
-
-
|
|
124
|
+
- Figma-supported and skilly-hand-supported via client docs/catalog: `gemini`.
|
|
125
|
+
- Figma-supported but not skilly-hand-native: Amazon Q, Android Studio, VS Code, Warp, Augment, Factory, Firebender, Kiro, OpenHands, Replit.
|
|
126
|
+
- skilly-hand-supported but not source-backed in current Figma docs: `antigravity`, `windsurf`, `trae`.
|
|
@@ -5,6 +5,7 @@
|
|
|
5
5
|
- "Set up Figma MCP remote server and verify with `whoami`."
|
|
6
6
|
- "Use Figma MCP `whoami` and tell me which account and plans are active."
|
|
7
7
|
- "Check whether this client exposes the Figma MCP tools I need before starting."
|
|
8
|
+
- "Check whether my client has the Figma-provided skills installed before starting a write-to-canvas workflow."
|
|
8
9
|
|
|
9
10
|
## Read Context
|
|
10
11
|
|
|
@@ -22,12 +23,14 @@
|
|
|
22
23
|
- "Use `get_code_connect_map` for this node and show mapped source locations."
|
|
23
24
|
- "Use `add_code_connect_map` to map this Figma node to my component path."
|
|
24
25
|
- "Use `get_code_connect_suggestions`, then confirm mappings with `send_code_connect_mappings` if the suggestions are correct."
|
|
26
|
+
- "Use `figma-create-design-system-rules` to generate project rules before implementing designs from Figma."
|
|
25
27
|
|
|
26
28
|
## First Canvas Creation
|
|
27
29
|
|
|
28
30
|
- "Use `whoami` to choose the right plan, then use `create_new_file` to create a new design file named 'MCP First Canvas'."
|
|
29
31
|
- "Use `create_new_file` to create a new design file named 'MCP First Canvas'."
|
|
30
32
|
- "Use `use_figma` to create one frame, add one title text layer, and apply auto layout spacing."
|
|
33
|
+
- "Use `figma-use` with `use_figma` for a small Figma Design write, then validate with a screenshot."
|
|
31
34
|
- "After the write, use `get_screenshot` to verify the result."
|
|
32
35
|
- "Before writing, inspect the file and use `search_design_system` for existing components or variables."
|
|
33
36
|
|
|
@@ -40,11 +43,20 @@
|
|
|
40
43
|
## FigJam Flows
|
|
41
44
|
|
|
42
45
|
- "Use `create_new_file` to create a new FigJam board for architecture planning."
|
|
46
|
+
- "Use `figma-create-new-file figjam 'Architecture Planning'` to create a new FigJam board."
|
|
43
47
|
- "Use `generate_diagram` to create a sequence diagram for login and checkout flow."
|
|
44
48
|
- "Use `generate_diagram` to add an ERD to this existing FigJam file."
|
|
45
|
-
- "Use `use_figma` to organize this FigJam board into sections with stickies and connectors."
|
|
49
|
+
- "Use `figma-use-figjam` with `use_figma` to organize this FigJam board into sections with stickies and connectors."
|
|
46
50
|
- "Use `use_figma` to update this architecture diagram with a new service."
|
|
47
51
|
|
|
52
|
+
## Figma Skills
|
|
53
|
+
|
|
54
|
+
- "Use `figma-use` for this Figma Design write-to-canvas task."
|
|
55
|
+
- "Use `figma-use-figjam` for this FigJam board update."
|
|
56
|
+
- "Use `figma-generate-design` to build this app screen in Figma using the connected design system."
|
|
57
|
+
- "Use `figma-generate-library` to create or sync a Figma design-system library from this codebase."
|
|
58
|
+
- "Use `figma-implement-design` to turn this Figma frame URL into production code."
|
|
59
|
+
|
|
48
60
|
## Agent Coverage
|
|
49
61
|
|
|
50
62
|
- "For this agent, tell me whether Figma officially supports its MCP setup path and which command/config I should use."
|
|
@@ -56,3 +68,4 @@
|
|
|
56
68
|
- "I am rate-limited. Switch to a staged workflow with fewer `get_design_context` calls."
|
|
57
69
|
- "This client does not expose `get_screenshot`. Choose the closest supported verification path."
|
|
58
70
|
- "My `create_new_file` call failed. Check whether a plan key or team context is missing."
|
|
71
|
+
- "Images from the MCP response expired. Refresh the design context or save the assets locally."
|
|
@@ -10,10 +10,10 @@
|
|
|
10
10
|
"agentSupport": ["codex", "claude", "cursor", "gemini", "copilot", "antigravity", "windsurf", "trae"],
|
|
11
11
|
"skillMetadata": {
|
|
12
12
|
"author": "skilly-hand",
|
|
13
|
-
"last-edit": "2026-05-
|
|
13
|
+
"last-edit": "2026-05-09",
|
|
14
14
|
"license": "Apache-2.0",
|
|
15
|
-
"version": "1.0.
|
|
16
|
-
"changelog": "
|
|
15
|
+
"version": "1.0.3",
|
|
16
|
+
"changelog": "Updated Figma MCP guidance for current remote-first docs, supported-client coverage, Figma-provided skills, FigJam write workflows, code-to-canvas support, and access/rate-limit language; affects setup routing, tool selection, prompt assets, and official matrix",
|
|
17
17
|
"auto-invoke": "Installing, configuring, or using Figma MCP from setup through first canvas creation",
|
|
18
18
|
"allowed-modes": [
|
|
19
19
|
"install-auth",
|
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
# Official Figma MCP Tools Matrix (As of 2026-05-
|
|
1
|
+
# Official Figma MCP Tools Matrix (As of 2026-05-09)
|
|
2
2
|
|
|
3
3
|
This matrix tracks the official Figma MCP tool/function set in current Figma documentation. Treat it as the source of truth for official tool names, supported file types, and remote-only status. Client-specific helpers are listed separately.
|
|
4
4
|
|
|
@@ -19,7 +19,7 @@ This matrix tracks the official Figma MCP tool/function set in current Figma doc
|
|
|
19
19
|
| `whoami` | Show authenticated Figma identity, plans, and seat types | None | No file context required | Remote-only in official docs; exempt from read rate limits. |
|
|
20
20
|
| `get_code_connect_suggestions` | Suggest candidate node-to-code mappings | File context + repo context | Figma Design | Usually part of Code Connect workflows. |
|
|
21
21
|
| `send_code_connect_mappings` | Confirm/persist mapping suggestions | Suggestions payload | Figma Design | Follow-up action after suggestion generation. |
|
|
22
|
-
| `use_figma` | General-purpose write/edit/delete/inspect via Plugin API workflow | Task instructions and file context | Figma Design, FigJam | Remote-only; default direct write/update path;
|
|
22
|
+
| `use_figma` | General-purpose write/edit/delete/inspect via Plugin API workflow | Task instructions and file context | Figma Design, FigJam | Remote-only; default direct write/update path; use `figma-use` for Figma Design and `figma-use-figjam` for FigJam when available. |
|
|
23
23
|
| `search_design_system` | Search libraries for components, variables, styles | Query text and optional type narrowing | Figma Design | Remote-only; use before creating new artifacts when design-system reuse is relevant. |
|
|
24
24
|
| `create_new_file` | Create new blank Design or FigJam file | File name, file type, plan context | No file context required | Remote-only; creates files in drafts or a project; requires plan/team context in some clients. |
|
|
25
25
|
|
|
@@ -27,6 +27,21 @@ This matrix tracks the official Figma MCP tool/function set in current Figma doc
|
|
|
27
27
|
|
|
28
28
|
Figma Make project context is exposed through MCP resources on clients that support MCP resources. Use a valid Figma Make link, let the client list available project files, and fetch only the files needed for the implementation task.
|
|
29
29
|
|
|
30
|
+
## Figma-Provided Skills
|
|
31
|
+
|
|
32
|
+
Figma documents these skills as the preferred guidance layer for common MCP workflows:
|
|
33
|
+
|
|
34
|
+
| Skill | Purpose |
|
|
35
|
+
| --- | --- |
|
|
36
|
+
| `figma-use` | Foundational Figma Design write-to-canvas workflow. |
|
|
37
|
+
| `figma-use-figjam` | Foundational FigJam write workflow. |
|
|
38
|
+
| `figma-code-connect-components` | Connect published Figma components to code implementations. |
|
|
39
|
+
| `figma-create-design-system-rules` | Generate project rules for design-to-code workflows. |
|
|
40
|
+
| `figma-create-new-file` | Create blank Design or FigJam files. |
|
|
41
|
+
| `figma-implement-design` | Translate a Figma design URL into production code. |
|
|
42
|
+
| `figma-generate-library` | Example workflow for generating or syncing a design-system library. |
|
|
43
|
+
| `figma-generate-design` | Example workflow for building screens/views from a design system. |
|
|
44
|
+
|
|
30
45
|
## Codex Plugin Helpers
|
|
31
46
|
|
|
32
47
|
The Codex Figma plugin may expose helper tools beyond the current official tools page. Keep these documented as client-specific helpers:
|
|
@@ -41,16 +56,19 @@ Some official Figma MCP tools may not be exposed by every client plugin surface.
|
|
|
41
56
|
|
|
42
57
|
| Coverage | Agents or Clients | Guidance |
|
|
43
58
|
| --- | --- | --- |
|
|
44
|
-
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot` | Include concrete setup paths
|
|
45
|
-
| Figma-supported but not skilly-hand-native | VS Code, Warp, Augment, Factory, Firebender | Mention as Figma-supported clients; do not add skilly-hand install assumptions. |
|
|
46
|
-
| skilly-hand-supported but not source-backed in current Figma docs | `
|
|
59
|
+
| Figma-supported and skilly-hand-supported | `codex`, `claude`, `cursor`, `copilot`, `gemini` | Include concrete setup paths where documented; otherwise point to the Figma MCP catalog/client docs. |
|
|
60
|
+
| Figma-supported but not skilly-hand-native | Amazon Q, Android Studio, VS Code, Warp, Augment, Factory, Firebender, Kiro, OpenHands, Replit | Mention as Figma-supported clients; do not add skilly-hand install assumptions. |
|
|
61
|
+
| skilly-hand-supported but not source-backed in current Figma docs | `antigravity`, `windsurf`, `trae` | Keep broad install support; advise manual MCP/client docs for Figma-specific setup. |
|
|
47
62
|
|
|
48
63
|
## Rate-Limit Notes (Official)
|
|
49
64
|
|
|
50
65
|
- Read tools are subject to seat/plan limits and per-minute limits.
|
|
66
|
+
- Remote server access is documented for all seats/plans, but practical usage is governed by seat/plan limits.
|
|
67
|
+
- Desktop server access is documented for Dev or Full seats on paid plans.
|
|
51
68
|
- Starter plan or View/Collab seats: up to 6 tool calls per month.
|
|
52
69
|
- Professional or Organization with Dev/Full seat: up to 200 tool calls per day.
|
|
53
70
|
- Enterprise with Dev/Full seat: up to 600 tool calls per day.
|
|
71
|
+
- Dev seats only have read-only access outside drafts for write-to-canvas workflows.
|
|
54
72
|
- Officially documented exempt tools include:
|
|
55
73
|
- `add_code_connect_map`
|
|
56
74
|
- `generate_figma_design`
|
|
@@ -65,6 +83,7 @@ Some official Figma MCP tools may not be exposed by every client plugin surface.
|
|
|
65
83
|
5. Use `use_figma` for direct Figma Design or FigJam writes and updates.
|
|
66
84
|
6. Keep `use_figma` calls small and validate after each step.
|
|
67
85
|
7. Use `get_screenshot` as final visual verification after edits when available.
|
|
86
|
+
8. Use `figma-use-figjam` rather than Figma Design assumptions for FigJam board edits.
|
|
68
87
|
|
|
69
88
|
## Sources
|
|
70
89
|
|
|
@@ -77,3 +96,4 @@ Some official Figma MCP tools may not be exposed by every client plugin surface.
|
|
|
77
96
|
- https://developers.figma.com/docs/figma-mcp-server/bringing-make-context-to-your-agent/
|
|
78
97
|
- https://developers.figma.com/docs/figma-mcp-server/plans-access-and-permissions/
|
|
79
98
|
- https://help.figma.com/hc/en-us/articles/32132100833559-Guide-to-the-Figma-MCP-server
|
|
99
|
+
- https://help.figma.com/hc/en-us/articles/39166810751895-Figma-skills-for-MCP
|
|
@@ -1,12 +1,12 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: "frontend-design"
|
|
3
|
-
description: "Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish."
|
|
3
|
+
description: "Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, taste-reference extraction, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish."
|
|
4
4
|
skillMetadata:
|
|
5
5
|
author: "skilly-hand"
|
|
6
|
-
last-edit: "2026-05-
|
|
6
|
+
last-edit: "2026-05-09"
|
|
7
7
|
license: "Apache-2.0"
|
|
8
|
-
version: "1.
|
|
9
|
-
changelog: "Added
|
|
8
|
+
version: "1.5.0"
|
|
9
|
+
changelog: "Added taste-reference extraction guidance sourced from Refero Styles and Impeccable; improves how agents translate visual references, anti-references, and register into DESIGN.md-ready language; affects greenfield design setup, critique, refinement, and resource routing"
|
|
10
10
|
auto-invoke: "Designing or generating UI components, pages, or layouts in a web or mobile project; setting up visual direction for a greenfield project; critiquing generated UI for AI slop; adding motion or micro-interactions to existing UI; refining or polishing generated UI output"
|
|
11
11
|
allowed-tools:
|
|
12
12
|
- "Read"
|
|
@@ -59,11 +59,12 @@ Always run stack detection first. Never skip to design.
|
|
|
59
59
|
2. **Check for DESIGN.md** — if it exists, read it before any design work. If it does not exist and the project has no existing components to sample, run `design-context-setter` to create it.
|
|
60
60
|
3. **Present findings to the user** — surface the detected stack and any DESIGN.md context clearly, then ask for explicit confirmation.
|
|
61
61
|
4. **If anything is unclear or ambiguous, ask** — do not proceed with partial or uncertain information.
|
|
62
|
-
5. **
|
|
63
|
-
6. **
|
|
64
|
-
7. **
|
|
65
|
-
8. **
|
|
66
|
-
9. **
|
|
62
|
+
5. **Extract taste from references** — when the user provides visual references or asks for stronger taste, use [assets/taste-reference-extraction.md](assets/taste-reference-extraction.md) to translate examples into concrete design language, anti-references, and register.
|
|
63
|
+
6. **Scan existing tokens and components** — read what already exists before proposing anything.
|
|
64
|
+
7. **Design with confirmed context only** — hand off to `component-designer` only after steps 2–5 are complete.
|
|
65
|
+
8. **Critique after generation** — invoke `critique` for a frontend-only challenge pass before polish.
|
|
66
|
+
9. **Refine from critique** — invoke `visual-refiner` for visual fixes routed by critique.
|
|
67
|
+
10. **Optionally add motion** — invoke `motion-designer` if critique, refinement, or the user identifies a motion need. Route Motion-native JavaScript/React animation through `motion-animation`; route GSAP timelines, ScrollTrigger, and plugin decisions through `gsap-animation`.
|
|
67
68
|
|
|
68
69
|
---
|
|
69
70
|
|
|
@@ -136,6 +137,17 @@ Every new component or style must feel like it was written by the same team that
|
|
|
136
137
|
|
|
137
138
|
If no existing components are found, use `DESIGN.md` as the visual language reference. If neither exists, run `design-context-setter` before proceeding.
|
|
138
139
|
|
|
140
|
+
### Pattern 5: Explain Taste as Observable Decisions
|
|
141
|
+
|
|
142
|
+
When the user provides references, do not summarize them as vibes. Convert each reference into visible, buildable decisions:
|
|
143
|
+
|
|
144
|
+
- **Register:** brand surface where design is the product, or product surface where design serves repeated use.
|
|
145
|
+
- **Visual ingredients:** type contrast, color role, spacing rhythm, density, radius, elevation, imagery, component shape, and motion character.
|
|
146
|
+
- **Taste rules:** what to repeat, what to avoid, and what would make the design feel off-brand.
|
|
147
|
+
- **Anti-references:** common AI reflexes the project should reject.
|
|
148
|
+
|
|
149
|
+
Use [assets/taste-reference-extraction.md](assets/taste-reference-extraction.md) for the extraction workflow. Its source model combines Refero Styles' reference-search framing with Impeccable's design vocabulary, register split, and anti-slop detection.
|
|
150
|
+
|
|
139
151
|
---
|
|
140
152
|
|
|
141
153
|
## What Not To Do
|
|
@@ -290,3 +302,4 @@ find src/components -maxdepth 2 -name "*.tsx" -o -name "*.vue" | head -10
|
|
|
290
302
|
- Motion and micro-interactions: [agents/motion-designer.md](agents/motion-designer.md)
|
|
291
303
|
- Full scan checklist: [assets/stack-scan-checklist.md](assets/stack-scan-checklist.md)
|
|
292
304
|
- Aesthetic archetypes reference: [assets/aesthetic-archetypes.md](assets/aesthetic-archetypes.md)
|
|
305
|
+
- Taste reference extraction: [assets/taste-reference-extraction.md](assets/taste-reference-extraction.md)
|
|
@@ -53,6 +53,7 @@ These apply within the constraints of the confirmed stack — they guide choices
|
|
|
53
53
|
- One clear visual focus per component — the most important element should be unmistakable.
|
|
54
54
|
- Avoid equal-weight elements. If everything is emphasized, nothing is.
|
|
55
55
|
- When `DESIGN.md` defines a motion character, let it inform transition presence even at this stage — flag it for `motion-designer` rather than omitting it entirely.
|
|
56
|
+
- If `DESIGN.md` includes visual references or anti-references, translate them through [../assets/taste-reference-extraction.md](../assets/taste-reference-extraction.md): borrow observable decisions, not full layouts.
|
|
56
57
|
|
|
57
58
|
---
|
|
58
59
|
|
|
@@ -12,6 +12,7 @@ Before critiquing:
|
|
|
12
12
|
2. Read `DESIGN.md` if it exists.
|
|
13
13
|
3. Inspect the generated/proposed UI source and, when available, the rendered page.
|
|
14
14
|
4. Compare against sampled project components, tokens, and interaction patterns.
|
|
15
|
+
5. If `DESIGN.md` includes references or anti-references, evaluate against the taste rules from [../assets/taste-reference-extraction.md](../assets/taste-reference-extraction.md), not just generic taste.
|
|
15
16
|
|
|
16
17
|
Do not use this agent when:
|
|
17
18
|
|
|
@@ -65,6 +66,8 @@ Check whether the UI serves the actual product context.
|
|
|
65
66
|
- Brand: Does it follow `DESIGN.md` personality and anti-references?
|
|
66
67
|
- Mode: Is this a brand surface where design is the product, or a product surface where design serves repeated use?
|
|
67
68
|
- Distinction: Would this still be recognizable if the logo and copy were removed?
|
|
69
|
+
- Reference translation: Did the UI borrow concrete ingredients from the references, or did it imitate a superficial mood?
|
|
70
|
+
- Anti-reference compliance: Did any explicitly rejected pattern slip back in?
|
|
68
71
|
|
|
69
72
|
### Pass 3 - Nielsen Heuristics
|
|
70
73
|
|
|
@@ -6,6 +6,8 @@ Gather the project's design intent once, then write it to `DESIGN.md` at the pro
|
|
|
6
6
|
|
|
7
7
|
This agent is modeled on how modern AI-first design platforms (Stitch, v0, Galileo) treat a persistent design brief — a short, always-available document that anchors every generation.
|
|
8
8
|
|
|
9
|
+
When the user provides references, use the extraction workflow in [../assets/taste-reference-extraction.md](../assets/taste-reference-extraction.md) before drafting the aesthetic direction. It converts reference sites, brands, moods, and anti-references into concrete language the rest of the skill can apply.
|
|
10
|
+
|
|
9
11
|
---
|
|
10
12
|
|
|
11
13
|
## When to Use
|
|
@@ -59,10 +61,20 @@ Ask these questions one at a time. Do not front-load the full list.
|
|
|
59
61
|
**4. Are there any visual references?**
|
|
60
62
|
"Any products, sites, or brands whose aesthetic this should feel close to? (Optional — skip if none.)"
|
|
61
63
|
|
|
62
|
-
|
|
64
|
+
If the user provides references, extract them into:
|
|
65
|
+
|
|
66
|
+
- what to borrow;
|
|
67
|
+
- what to avoid;
|
|
68
|
+
- whether the target is brand or product register;
|
|
69
|
+
- the visible ingredients that define the taste.
|
|
70
|
+
|
|
71
|
+
**5. Any anti-references?**
|
|
72
|
+
"Are there products, sites, or AI-looking patterns this should explicitly avoid? Examples: purple gradients, glassmorphism, nested cards, generic centered hero layouts."
|
|
73
|
+
|
|
74
|
+
**6. What's the accessibility baseline?**
|
|
63
75
|
"Should we target WCAG 2.2 Level AA (standard), AAA (enhanced), or is there a specific requirement? Default is AA."
|
|
64
76
|
|
|
65
|
-
**
|
|
77
|
+
**7. Any hard constraints?**
|
|
66
78
|
"Are there colors, fonts, or patterns that must be used or must be avoided? (brand guidelines, corporate requirements, legal restrictions)"
|
|
67
79
|
|
|
68
80
|
After collecting answers, propose a brief aesthetic direction and ask for confirmation before writing the file.
|
|
@@ -89,6 +101,7 @@ Write the following structure to `DESIGN.md` at the project root. Every field mu
|
|
|
89
101
|
|
|
90
102
|
**Adjectives:** [3 adjectives from user]
|
|
91
103
|
**Visual references:** [references, or "none specified"]
|
|
104
|
+
**Anti-references:** [visual patterns or references to avoid, or "none specified"]
|
|
92
105
|
|
|
93
106
|
## Aesthetic Direction
|
|
94
107
|
|
|
@@ -35,6 +35,8 @@ Run all four checks. Do not skip any. Report findings grouped by check — do no
|
|
|
35
35
|
|
|
36
36
|
Look for generic patterns that signal uncontextualized AI output. Flag each one found.
|
|
37
37
|
|
|
38
|
+
If `DESIGN.md` includes visual references or anti-references, run the reference taste check in [../assets/taste-reference-extraction.md](../assets/taste-reference-extraction.md) first. A design can be clean and still fail if it ignores the intended register, borrows the wrong ingredient from a reference, or drifts into an explicit anti-reference.
|
|
39
|
+
|
|
38
40
|
**Visual tells to catch:**
|
|
39
41
|
|
|
40
42
|
- Glassmorphism (backdrop-filter: blur + semi-transparent backgrounds with no established project precedent)
|
|
@@ -45,6 +47,8 @@ Look for generic patterns that signal uncontextualized AI output. Flag each one
|
|
|
45
47
|
- Gradient text (`background-clip: text`) used decoratively without project precedent
|
|
46
48
|
- Icon + title + description card grids as default empty-state filler
|
|
47
49
|
- Identical padding and border-radius across every element (uniform blandness)
|
|
50
|
+
- Product UI treated like a marketing page, or brand/landing work flattened into utilitarian dashboard language
|
|
51
|
+
- Reference mimicry that copies a layout while missing its real taste ingredients
|
|
48
52
|
|
|
49
53
|
For each flag: name the pattern, show the line, and suggest a project-derived alternative.
|
|
50
54
|
|
|
@@ -0,0 +1,161 @@
|
|
|
1
|
+
# Taste Reference Extraction
|
|
2
|
+
|
|
3
|
+
Use this when a user provides visual references, asks for better taste, or wants the agent to explain a design direction in language another AI can reliably apply.
|
|
4
|
+
|
|
5
|
+
This workflow is informed by two public reference systems:
|
|
6
|
+
|
|
7
|
+
- Refero Styles: https://styles.refero.design/
|
|
8
|
+
- Impeccable: https://impeccable.style/
|
|
9
|
+
|
|
10
|
+
Refero's useful pattern is searchable real-product reference extraction: brand, mood, color, typography, URL, spacing, components, and DESIGN.md-ready output. Impeccable's useful pattern is shared design vocabulary: product-vs-brand register, anti-references, command-like refinement dimensions, and deterministic anti-slop checks.
|
|
11
|
+
|
|
12
|
+
Do not copy either source's site language into project briefs. Use them as a model for how to structure taste into observable decisions.
|
|
13
|
+
|
|
14
|
+
---
|
|
15
|
+
|
|
16
|
+
## When to Use
|
|
17
|
+
|
|
18
|
+
Use this asset when:
|
|
19
|
+
|
|
20
|
+
- The user names sites, products, brands, moods, or URLs as references.
|
|
21
|
+
- `DESIGN.md` needs stronger language than vague adjectives.
|
|
22
|
+
- A generated UI feels generic but the fix is not obvious from tokens alone.
|
|
23
|
+
- The user asks to make something feel more premium, editorial, playful, precise, calm, technical, or distinctive.
|
|
24
|
+
|
|
25
|
+
Do not use it to override existing project tokens, accessibility requirements, or explicit user constraints.
|
|
26
|
+
|
|
27
|
+
---
|
|
28
|
+
|
|
29
|
+
## Extraction Ladder
|
|
30
|
+
|
|
31
|
+
Move from vague taste to buildable decisions in this order.
|
|
32
|
+
|
|
33
|
+
### 1. Register
|
|
34
|
+
|
|
35
|
+
Classify the surface before choosing visual moves.
|
|
36
|
+
|
|
37
|
+
| Register | Design Role | Good Taste Looks Like |
|
|
38
|
+
| --- | --- | --- |
|
|
39
|
+
| Brand | Design is the product: landing pages, launches, portfolios, editorial, venues, campaigns | Distinctive imagery, stronger type voice, memorable composition, carefully chosen drama |
|
|
40
|
+
| Product | Design serves repeated work: dashboards, tools, CRM, admin, commerce flows | Fast scanning, durable hierarchy, predictable controls, restrained styling, task-first density |
|
|
41
|
+
| Hybrid | Brand expression wraps a functional flow: onboarding, pricing, checkout, product-led homepage | Brand moments at entry points; product clarity at decision and action points |
|
|
42
|
+
|
|
43
|
+
Write the register into `DESIGN.md` when it is known.
|
|
44
|
+
|
|
45
|
+
### 2. Reference Inventory
|
|
46
|
+
|
|
47
|
+
For each reference, extract the visible ingredients:
|
|
48
|
+
|
|
49
|
+
```text
|
|
50
|
+
Reference:
|
|
51
|
+
Role: close model / partial ingredient / anti-reference
|
|
52
|
+
Register: brand / product / hybrid
|
|
53
|
+
Borrow:
|
|
54
|
+
- Typography:
|
|
55
|
+
- Color:
|
|
56
|
+
- Spacing and density:
|
|
57
|
+
- Layout structure:
|
|
58
|
+
- Component shape:
|
|
59
|
+
- Imagery or texture:
|
|
60
|
+
- Motion:
|
|
61
|
+
Avoid:
|
|
62
|
+
- Surface-level mimicry:
|
|
63
|
+
- Patterns that would not fit this product:
|
|
64
|
+
```
|
|
65
|
+
|
|
66
|
+
If the user provides only a mood, ask for one concrete reference or translate the mood into likely ingredients and mark them as assumptions.
|
|
67
|
+
|
|
68
|
+
### 3. Taste Rules
|
|
69
|
+
|
|
70
|
+
Convert references into rules the implementation can check.
|
|
71
|
+
|
|
72
|
+
Good taste rules are concrete:
|
|
73
|
+
|
|
74
|
+
- "Use color as state and emphasis, not as section decoration."
|
|
75
|
+
- "Favor one strong typographic contrast over multiple decorative containers."
|
|
76
|
+
- "Keep dashboard panels dense and flat; reserve depth for dialogs and active overlays."
|
|
77
|
+
- "Use asymmetry in hero composition, but keep form controls conventional."
|
|
78
|
+
|
|
79
|
+
Weak taste rules are vague:
|
|
80
|
+
|
|
81
|
+
- "Make it premium."
|
|
82
|
+
- "Use modern design."
|
|
83
|
+
- "Make it clean."
|
|
84
|
+
- "Add personality."
|
|
85
|
+
|
|
86
|
+
### 4. Anti-References
|
|
87
|
+
|
|
88
|
+
Name what must not happen. Anti-references prevent the model from falling back to common AI reflexes.
|
|
89
|
+
|
|
90
|
+
Capture:
|
|
91
|
+
|
|
92
|
+
- explicit sites or products to avoid;
|
|
93
|
+
- visual tropes to reject;
|
|
94
|
+
- category cliches to resist;
|
|
95
|
+
- font, palette, layout, or motion reflexes that would make the output generic.
|
|
96
|
+
|
|
97
|
+
Common anti-reference language:
|
|
98
|
+
|
|
99
|
+
- no purple-to-blue gradients unless already in the brand system;
|
|
100
|
+
- no glass cards, glow borders, or blurred background blobs as decoration;
|
|
101
|
+
- no nested card stacks for page sections;
|
|
102
|
+
- no icon-tile feature grid as the default explanation pattern;
|
|
103
|
+
- no centered hero plus metrics unless the product actually needs that story;
|
|
104
|
+
- no monospace as shorthand for technical credibility;
|
|
105
|
+
- no motion that does not communicate state, cause, or hierarchy.
|
|
106
|
+
|
|
107
|
+
### 5. Taste Contract
|
|
108
|
+
|
|
109
|
+
End with a short contract that can be pasted into `DESIGN.md` or a task prompt.
|
|
110
|
+
|
|
111
|
+
```markdown
|
|
112
|
+
## Taste Contract
|
|
113
|
+
|
|
114
|
+
**Register:** [brand/product/hybrid]
|
|
115
|
+
**Reference role:** [what each reference contributes]
|
|
116
|
+
**Borrow:** [3-5 concrete ingredients]
|
|
117
|
+
**Avoid:** [3-5 anti-references]
|
|
118
|
+
**Hierarchy:** [how attention should move]
|
|
119
|
+
**Density:** [sparse, balanced, dense, or context-specific]
|
|
120
|
+
**Color role:** [brand, state, emphasis, data, or restraint]
|
|
121
|
+
**Typography role:** [voice, clarity, editorial contrast, utility]
|
|
122
|
+
**Motion role:** [none, state feedback, spatial continuity, delight]
|
|
123
|
+
**Failure mode:** [what would make this feel generic or wrong]
|
|
124
|
+
```
|
|
125
|
+
|
|
126
|
+
---
|
|
127
|
+
|
|
128
|
+
## Reference Use Rules
|
|
129
|
+
|
|
130
|
+
- Borrow decisions, not whole layouts.
|
|
131
|
+
- Prefer real product evidence over mood words.
|
|
132
|
+
- Keep brand and product registers separate; do not critique one by the other.
|
|
133
|
+
- When references conflict, decide which ingredient each one owns.
|
|
134
|
+
- Let anti-references constrain reflexes before choosing new decoration.
|
|
135
|
+
- Translate every aesthetic claim into at least one implementation lever: type, color, spacing, density, shape, imagery, motion, or copy voice.
|
|
136
|
+
- Do not introduce new tokens without stack confirmation and user approval.
|
|
137
|
+
|
|
138
|
+
---
|
|
139
|
+
|
|
140
|
+
## Example
|
|
141
|
+
|
|
142
|
+
```text
|
|
143
|
+
User says: "Make it feel like Linear and Apple, but not another dark SaaS dashboard."
|
|
144
|
+
|
|
145
|
+
Register: product
|
|
146
|
+
Reference roles:
|
|
147
|
+
- Linear: density, command-center clarity, restrained dark surfaces
|
|
148
|
+
- Apple: precision, generous detail spacing, high trust, fewer competing borders
|
|
149
|
+
Borrow:
|
|
150
|
+
- crisp text hierarchy with clear primary action
|
|
151
|
+
- muted surfaces with state-driven accent color
|
|
152
|
+
- shallow depth and precise separators instead of card stacks
|
|
153
|
+
- dense tables, but roomy detail panes
|
|
154
|
+
Avoid:
|
|
155
|
+
- neon-on-dark glow effects
|
|
156
|
+
- purple gradients
|
|
157
|
+
- giant centered hero metrics
|
|
158
|
+
- generic rounded cards around every group
|
|
159
|
+
Failure mode:
|
|
160
|
+
- looking like a launch page instead of a work surface
|
|
161
|
+
```
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"id": "frontend-design",
|
|
3
3
|
"title": "Frontend Design",
|
|
4
|
-
"description": "Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish.",
|
|
4
|
+
"description": "Project-aware frontend design skill that detects the existing tech stack, UI libraries, CSS variables, and design tokens before proposing any UI work. Supports greenfield projects via DESIGN.md context setup, taste-reference extraction, post-generation critique, visual refinement, and Motion/GSAP-aware motion polish.",
|
|
5
5
|
"portable": true,
|
|
6
6
|
"tags": ["frontend", "design", "workflow", "ui", "motion", "greenfield"],
|
|
7
7
|
"detectors": ["always"],
|
|
@@ -10,10 +10,10 @@
|
|
|
10
10
|
"agentSupport": ["codex", "claude", "cursor", "gemini", "copilot", "antigravity", "windsurf", "trae"],
|
|
11
11
|
"skillMetadata": {
|
|
12
12
|
"author": "skilly-hand",
|
|
13
|
-
"last-edit": "2026-05-
|
|
13
|
+
"last-edit": "2026-05-09",
|
|
14
14
|
"license": "Apache-2.0",
|
|
15
|
-
"version": "1.
|
|
16
|
-
"changelog": "Added
|
|
15
|
+
"version": "1.5.0",
|
|
16
|
+
"changelog": "Added taste-reference extraction guidance sourced from Refero Styles and Impeccable; improves how agents translate visual references, anti-references, and register into DESIGN.md-ready language; affects greenfield design setup, critique, refinement, and resource routing",
|
|
17
17
|
"auto-invoke": "Designing or generating UI components, pages, or layouts in a web or mobile project; setting up visual direction for a greenfield project; critiquing generated UI for AI slop; adding motion or micro-interactions to existing UI; refining or polishing generated UI output",
|
|
18
18
|
"allowed-modes": [
|
|
19
19
|
"stack-detector",
|
|
@@ -34,7 +34,8 @@
|
|
|
34
34
|
{ "path": "agents/visual-refiner.md", "kind": "asset" },
|
|
35
35
|
{ "path": "agents/motion-designer.md", "kind": "asset" },
|
|
36
36
|
{ "path": "assets/stack-scan-checklist.md", "kind": "asset" },
|
|
37
|
-
{ "path": "assets/aesthetic-archetypes.md", "kind": "asset" }
|
|
37
|
+
{ "path": "assets/aesthetic-archetypes.md", "kind": "asset" },
|
|
38
|
+
{ "path": "assets/taste-reference-extraction.md", "kind": "asset" }
|
|
38
39
|
],
|
|
39
40
|
"dependencies": ["motion-animation", "gsap-animation"]
|
|
40
41
|
}
|
package/package.json
CHANGED