How to use
- Copy the installation command and review the source package before installing it in your own agent environment.
- Check the source document for required tools, credentials and supported hosts.
- Describe your input and desired deliverable; review the output against the source material.
npx skills add https://github.com/magentosh/superpowers --skill geminiBefore you start
- Installation does not include a model subscription, API allowance or permission to access your files.
- Repository stars and installation counts describe activity, not an independent quality or security test.
Rendered from the source repository. Headings, examples, code, tables, links, and referenced images are preserved.
Install the belt CLI skill: npx skills add belt-sh/cliGemini
Run Google's Gemini chat models via inference.sh CLI.
Quick Start
Requires inference.sh CLI (belt). Install instructionsbelt login
belt app run google/gemini-3-8-flash --input '{"text": "Explain how HTTPS certificate validation works"}'Models
| Model | App ID | Best For |
|---|---|---|
| Gemini 4 Argon | google/gemini-4-argon | Flagship: deep multi-step work in coding, finance, law, security, science |
| Gemini 3.8 Flash | google/gemini-3-8-flash | Most capable Flash: long-horizon software engineering, autonomous agents, enterprise workflows |
| Gemini 3.5 Flash-Lite | google/gemini-3-5-flash-lite | Fastest, most cost-effective 3.5-class model (350 output tokens/s), high-throughput agentic workflows |
All run via Vertex AI with vision and tool use. Argon and 3.8 Flash have a 1M token context window; Argon supports thinking levels low to high.
Other Gemini apps (older chat versions, image models, and the Gemini Omni video models): belt app search gemini. google/gemini-omni-1-2-flash is a video generation model, not a chat model.
Inputs
| Field | Description | |||
|---|---|---|---|---|
text | The user message | |||
system_prompt | System instructions | |||
images | Image files or URLs (vision) | |||
files | Document files | |||
context | Prior messages: `[{"role": "user" \ | "assistant", "text": "..."}]` | ||
reasoning_effort | none (default), low, medium, high | |||
temperature | Default 1.0 | |||
top_k | Default 40, -1 to disable | |||
max_tokens | Default 64000 | |||
tools | Function tool definitions | |||
tool_choice | `{"mode": "auto" \ | "none" \ | "required" \ | "function", "name": ...}` |
response_format | `{"type": "text" \ | "json_object" \ | "json_schema", ...}` |
Output: response, reasoning, tool_calls, usage.
Each app also has an openai function that takes Chat Completions messages: belt app sample google/gemini-4-argon -f openai.
Examples
Deep Analysis with Argon
belt app run google/gemini-4-argon --input '{
"system_prompt": "You are a securities lawyer. Cite the clause for each finding.",
"text": "Review this term sheet for investor-unfriendly provisions",
"files": ["https://example.com/term-sheet.pdf"],
"reasoning_effort": "high"
}'Agentic Coding with 3.8 Flash
belt app run google/gemini-3-8-flash --input '{
"system_prompt": "You are a Python engineer. Return a unified diff.",
"text": "Add retry with exponential backoff to this HTTP client: <code>",
"reasoning_effort": "medium"
}'High Throughput with Flash-Lite
belt app run google/gemini-3-5-flash-lite --input '{
"text": "Translate to Spanish: The meeting is moved to Thursday.",
"reasoning_effort": "none",
"max_tokens": 100
}'Run many inputs at once:
# one JSON input per line
belt app run google/gemini-3-5-flash-lite --batch inputs.jsonlVision
belt app run google/gemini-3-8-flash --input '{
"text": "Describe this chart and list the three largest values",
"images": ["https://example.com/chart.png"]
}'Multi-turn Conversation
belt app run google/gemini-4-argon --input '{
"context": [
{"role": "user", "text": "I am planning a 5-day trip to Kyoto."},
{"role": "assistant", "text": "Great choice. What are your interests?"}
],
"text": "Temples and food. Draft an itinerary."
}'Structured Output
belt app run google/gemini-3-5-flash-lite --input '{
"text": "Extract product and price: The Pixel case costs $29.",
"response_format": {
"type": "json_schema",
"name": "product",
"json_schema": {
"type": "object",
"properties": {"product": {"type": "string"}, "price": {"type": "number"}},
"required": ["product", "price"]
}
}
}'Tool Use
belt app run google/gemini-3-8-flash --input '{
"text": "Find open issues labeled bug",
"tools": [{
"type": "function",
"function": {
"name": "search_issues",
"description": "Search the issue tracker",
"parameters": {"type": "object", "properties": {"label": {"type": "string"}, "state": {"type": "string"}}, "required": ["label"]}
}
}],
"tool_choice": {"mode": "auto"}
}'Pricing
| Model | Input | Output |
|---|---|---|
| Gemini 4 Argon | $2.00/M | $10.00/M |
| Gemini 3.8 Flash | $1.50/M | $7.50/M |
| Gemini 3.5 Flash-Lite | $0.30/M | $2.50/M |
Estimate a run: belt app estimate <app-id> --input input.json.
Related Skills
# Full platform skill (all apps)
npx skills add inference-sh/skills@infsh-cli
# All LLMs in one place
npx skills add inference-sh/skills@llm-models
# Other providers
npx skills add inference-sh/skills@claude
npx skills add inference-sh/skills@gpt
# Agents and tool use
npx skills add inference-sh/skills@agent-tools
# RAG pipelines
npx skills add inference-sh/skills@ai-rag-pipelineBrowse all apps: belt app list
Documentation
- Agents Overview - Building AI agents
- Running Apps - How to run apps via CLI