veniceai · jonisjongithub · Apr 24, 2026
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -11,7 +11,19 @@ template/SKILL.md               Starting point for new skills
 
 Supporting files (helper scripts, reference data) can live alongside `SKILL.md` inside the skill folder and be referenced relative to it.
 
-## Style
+## Skill types
+
+This repo contains two types of skills:
+
+### API Documentation Skills (`venice-*`)
+
+Skills that teach agents how to use specific Venice API endpoints. These follow a strict template.
+
+### Generator / Scaffolding Skills (`create-*`)
+
+Skills that help agents create new projects using Venice AI. These include sample code, references, and interactive checklists. They follow a more flexible structure but must still have concrete frontmatter descriptions.
+
+## Style for API Documentation Skills
 
 - **Scope one surface area** — each skill covers a coherent slice of the Venice API (e.g. `venice-embeddings`, not `venice-embeddings-and-chat`). If two skills keep cross-referencing each other, either merge them or tighten their boundaries.
 - **Concrete frontmatter `description`** — the agent uses this to decide when to load. Name the specific endpoints, parameters, and scenarios. Vague descriptions ("things about chat") hurt selection.
@@ -21,6 +33,14 @@ Supporting files (helper scripts, reference data) can live alongside `SKILL.md`
 - **Cross-links** — reference related skills with relative paths (`../venice-errors/SKILL.md`).
 - **Length** — aim for under 500 lines. Longer skills should be split.
 
+## Style for Generator Skills
+
+- **Concrete frontmatter `description`** — describe when an agent should use this skill (e.g., "when building an agent TUI", "when scaffolding a new project").
+- **Interactive checklists** — present options as multi-select checklists with sensible defaults.
+- **Working sample code** — include a complete, runnable sample in a `sample/` directory.
+- **References** — put detailed specs in a `references/` directory to keep the main SKILL.md focused.
+- **Generation workflow** — clearly document the steps an agent should follow.
+
 ## Authoring a new skill
 
 1. Copy `template/` to `skills/<your-skill-name>/`.
@@ -38,10 +58,21 @@ Supporting files (helper scripts, reference data) can live alongside `SKILL.md`
 
 ## Review checklist
 
+### For API Documentation Skills
+
 - [ ] Frontmatter `name` + `description` present and concrete.
 - [ ] Endpoint table at the top.
 - [ ] At least one `curl` example.
 - [ ] Errors + gotchas section.
 - [ ] Cross-links to related skills.
 - [ ] Added to root `README.md` catalog if new.
 - [ ] No secrets or real API keys in examples (use `$VENICE_API_KEY`).
+
+### For Generator Skills
+
+- [ ] Frontmatter `name` + `description` present and concrete.
+- [ ] Interactive checklist with defaults marked.
+- [ ] Working sample code in `sample/` that compiles/runs.
+- [ ] Clear generation workflow documented.
+- [ ] Added to root `README.md` catalog if new.
+- [ ] No secrets or real API keys in examples (use `$VENICE_API_KEY`).
diff --git a/README.md b/README.md
@@ -13,6 +13,10 @@ template/      Copy this as a starting point for a new skill
 
 ## Skill catalog
 
+### API Documentation Skills
+
+Skills that teach agents how to use specific Venice API endpoints.
+
 | Skill | Covers |
 |---|---|
 | [`venice-api-overview`](./skills/venice-api-overview/SKILL.md) | Base URL, auth modes, response headers, pricing model, versioning |
@@ -35,6 +39,14 @@ template/      Copy this as a starting point for a new skill
 | [`venice-augment`](./skills/venice-augment/SKILL.md) | `/augment/text-parser`, `/augment/scrape`, `/augment/search` |
 | [`venice-errors`](./skills/venice-errors/SKILL.md) | Error shapes, 402 payment required, 422 content policy, 429 rate limits, retry strategy |
 
+### Generator / Scaffolding Skills
+
+Skills that help agents create new projects using Venice AI.
+
+| Skill | Covers |
+|---|---|
+| [`create-agent-tui`](./skills/create-agent-tui/SKILL.md) | Scaffold complete agent TUI projects in TypeScript — like `create-react-app` for terminal agents |
+
 ## Using these skills
 
 Each skill is just a folder with a `SKILL.md` that starts with YAML frontmatter:

diff --git a/skills/create-agent-tui/.gitignore b/skills/create-agent-tui/.gitignore
@@ -0,0 +1,5 @@
+node_modules/
+dist/
+.env
+*.log
+.DS_Store
diff --git a/skills/create-agent-tui/README.md b/skills/create-agent-tui/README.md
@@ -0,0 +1,93 @@
+# create-venice-agent-tui
+
+> Like `create-react-app` but for terminal agents — scaffolds a complete agent TUI in TypeScript using Venice AI.
+
+![Venice Agent TUI](./sample/screenshots/banner.png)
+
+## What It Does
+
+A skill for AI coding agents (Claude Code, Cursor, Codex, etc.) that scaffolds a complete agent TUI in TypeScript. Tell your coding agent what kind of agent you want, and it generates a runnable project that:
+
+- Works with **any Venice model** (Claude, GPT-5.5, Grok, Llama, DeepSeek, Mistral)
+- Provides a **fully customizable terminal interface** with streaming output
+- Includes **file operations, shell, grep, glob** tools out of the box
+- Supports **web search and X/Twitter search** via Venice's built-in capabilities
+- Tracks **token usage and session persistence**
+- Respects Venice's **privacy-first, zero-retention** inference
+
+## Quick Start
+
+### Try the sample:
+
+```bash
+cd sample
+npm install
+VENICE_API_KEY=vn_your_key npm start
+```
+
+### Or tell your coding agent:
+
+> "Build me an agent TUI that can read/write files and run shell commands"
+
+It will use this skill automatically.
+
+## Features
+
+### 🎨 Customizable UI
+- **Input styles:** block (background box), bordered (line frame), plain (simple prompt)
+- **Tool displays:** grouped (tree output), emoji (per-call markers), minimal (one-liners)
+- **Loaders:** spinner, gradient shimmer, trailing dots
+
+### 🛠️ Built-in Tools
+- File read/write/edit with diff validation
+- Shell command execution with timeout
+- Glob file finding
+- Grep content search
+- Custom tool templates
+
+### 🔍 Venice-Specific
+- Web search with citations
+- X/Twitter search (Grok models)
+- Reasoning effort control
+- DIEM cost tracking
+- Balance awareness
+
+### 📊 Session Management
+- JSONL conversation persistence
+- Token usage tracking
+- Multi-turn conversations
+- Slash commands (`/model`, `/new`, `/help`)
+
+## Model Aliases
+
+Quick shortcuts for common models:
+
+```
+opus     → claude-opus-4-6
+opus-4.7 → claude-opus-4-7
+sonnet   → claude-sonnet-4-6
+gpt5     → gpt-5.5
+grok     → grok-41-fast
+deepseek → deepseek-v4-pro
+llama    → llama-4-maverick-17b-128e-instruct
+qwen     → qwen3-235b-a22b
+mistral  → mistral-large-2411
+```
+
+Use with `/model grok` or `npm start -- --model=deepseek`.
+
+## Why Venice?
+
+1. **Privacy** — Zero data retention on private models
+2. **Model diversity** — 30+ models in one API
+3. **Built-in search** — Web and X search without extra API keys
+4. **DIEM economics** — Pay-as-you-go credits that can be resold
+5. **Uncensored options** — Models that work for legitimate use cases others refuse
+
+## License
+
+MIT
+
+---
+
+*Built for the Venice AI ecosystem. Get your API key at [venice.ai/settings/api](https://venice.ai/settings/api).*