agency-skill
Empowering Gemini to build premium digital experiences.
Open source Open in the app JSON README (API)
About
Empowering Gemini to build premium digital experiences.
Details
- Kind
- Plugins
- Topic
- AI, RAG & memory
- Publisher
- srinivas26k
- Origin
- gemini
- Category
- ferramentas
- Version
- 2.0.0
- Stars
- 4
- Forks
- 2
- Last push
- 2026-04-15T12:33:43Z
- Repository state
- ativo
- Language
- JavaScript
- License
- MIT
- Added
- 2026-08-30 14:13:39
- Updated
- 2026-08-30 14:13:39
- Origin id
srinivas26k/agency-skill
README
# agency-skill
**Agency Designer** is a cross-platform AI skill that brings agency-grade design, agent engineering, and conversion marketing into any AI IDE or coding environment.
Install it once. Use it in Claude, Copilot, Cursor, Gemini, Goose, Kiro, Windsurf — any environment that supports skills or instruction files.
[](https://www.npmjs.com/package/agency-skill)
[](LICENSE)
---
## What It Does
Three commands. Each one turns a vague idea into structured, production-ready output.
| Command | What it gives you |
|---|---|
| `/web-design` | A full design system — palette, typography, layout, motion, components — and a `design.md` your team can build from |
| `/agents-make` | A production-grade agent definition with model, system prompt, tools, MCP servers, skills, and callable agents — Anthropic Managed Agents style |
| `/marketing-pro` | Conversion-focused copy — headlines that follow the [Result without Pain] formula, specific CTAs, structured page copy, and an audit score |
Every command interrogates your brief one question at a time, selects the right sub-agents, presents a numbered plan, and waits for your approval before doing anything.
---
## Install
**Bootstrap into any project:**
```bash
npx agency-skill init
```
This copies the slash commands, agent definitions, and design resources directly into your project.
**Via Skills Registry:**
```bash
skills install agency-skill
```
**Via NPM:**
```bash
npm install agency-skill
```
---
## How It Works
1. Type `/web-design`, `/agents-make`, or `/marketing-pro` in your AI IDE
2. Answer a few brief questions — one at a time, no walls of text
3. Review the proposed plan
4. Approve it, then let the skill execute with the right agents in scope
---
## /web-design
Produces a complete design system for any product, page, or component.
**What you get:**
- Color palette as HSL CSS variables (psychology-first, not random)
- Premium font pairing with full type scale
- Layout structure and component language
- GSAP / Framer Motion animation layer
- A `design.md` file written to your project root — your living design system doc
**Sub-commands:**
```
/web-design component [name] → generate a single component
/web-design page [type] → full page (saas-landing, portfolio, agency, etc.)
/web-design palette [emotion] → psychology-first color system
/web-design font [personality] → premium font pairing
/web-design animate [type] → GSAP / Framer / Anime.js animation
/web-design audit → quality rubric score
/web-design style [1-100] → apply one of 100 graphic design styles
```
---
## /agents-make
Designs a real agent system — not a vague description of one.
Output follows the Anthropic Managed Agents definition format:
```json
{
"name": "Your Agent",
"model": "claude-sonnet-4-6",
"description": "What it does",
"system": "Behavioral rules and constraints",
"tools": [{ "type": "agent_toolset_20260401" }],
"mcp_servers": [],
"skills": [],
"callable_agents": [],
"metadata": {}
}
```
Plus: environment definition, session contract, and quality gates.
Works for any platform — if you're not on Anthropic Managed Agents, the same schema is adapted to your runtime.
**Sub-commands:**
```
/agents-make single → one production-ready agent definition
/agents-make pipeline → sequential pipeline with handoff contracts
/agents-make director → director + callable sub-agent tree
/agents-make managed → full Anthropic Managed Agents spec
/agents-make mcp → MCP server and tool boundary setup
/agents-make audit → review an existing system for gaps
```
---
## /marketing-pro
Structured conversion copy using proven frameworks.
**What you get:**
- Hero headline in 3 variants (using [Result without Pain] formula)
- Sub-headline and value proposition
- Primary and secondary CTAs — never "Learn More" or "Get Started"
- Specific social proof copy
- Section headlines and FAQ objection handlers
- Conversion audit score
**Frameworks available:** PAS, AIDA, FAB, BAB
**Sub-commands:**
```
/marketing-pro headline [context] → 5 headline variants
/marketing-pro cta [action] → CTA copy variants
/marketing-pro audit → conversion rubric score
/marketing-pro rewrite → full page copy rewrite
/marketing-pro email [type] → email sequence
/marketing-pro seo [keyword] → SEO-optimised copy
/marketing-pro ab [element] → A/B test variants
```
---
## Supported Environments
Works in any AI IDE or coding agent that supports skills, instruction files, or slash commands:
- Claude Code
- Claude Managed Agents
- GitHub Copilot
- Cursor
- Gemini
- Goose
- Kiro
- Windsurf
- VS Code agent workflows
- Any custom agent runtime
---
## Agents Included
The skill ships with five reusable agent role definitions your AI can invoke during any workflow:
- `@visual-director` — palette, typography, and style decisions
- `@motion-engineer` — GSAP, Framer Motion, and scroll behavior
- `@copy-strategist` — headlines, CTAs, and conversion copy
- `@quality-auditor` — S-Tier rubric check (scores 0–50 across typography, color, motion, imagery, copy)
- `@brief-analyst` — turns vague briefs into structured WHO/WHAT/FEEL/STYLE specs
---
## Anthropic Agent Docs Included
Full Anthropic Managed Agents documentation is bundled so your AI can reference the real spec while designing:
- Agent definition schema
- Tool configuration (built-in toolset + custom tools)
- MCP connector setup
- Skills attachment
- Environment and session planning
- Hallucination reduction patterns
---
## License
MIT. Created by [Srinivas Nampalli](https://www.linkedin.com/in/srinivas-nampalli/).