A cost-effective sub-agent that uses your Google Gemini subscription for high-context coding and text analysis within Openclaw Skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install gemini-sub-agent
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install gemini-sub-agent using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Gemini Sub-Agent?
The gemini-sub-agent allows developers to leverage their existing Google Gemini subscription to power sub-agent tasks without incurring additional API costs. It is specifically designed to handle medium-complexity tasks, long-context analysis, and autonomous coding workflows, making it a powerful addition to the library of Openclaw Skills.
By utilizing the Gemini CLI and gcloud authentication, this skill provides a bridge for running agentic tasks that would otherwise consume expensive API credits. It excels at processing large documents up to 1 million tokens and can autonomously read and write files during development, serving as a reliable alternative to paid cloud models for specific use cases.
Gemini Sub-Agent Use Cases
- Delegating medium-complexity coding tasks (30-200 lines) to save on Sonnet or Haiku credits.
- Processing and summarizing massive documents or logs using the 1M token context window.
- Running autonomous agentic coding sessions for refactoring or script generation with direct file access.
- Performing bulk text analysis, Q&A, and action item extraction from large reports or piped content.
How Gemini Sub-Agent Works
- The user authenticates their Google Gemini subscription via the gcloud CLI and Gemini CLI OAuth to enable access without API keys.
- The user invokes the ask-gemini wrapper for simple text tasks or the Gemini CLI for agentic workflows.
- For long-context tasks, content is piped directly into the skill, utilizing the massive context window of the Gemini models.
- In agentic mode, the skill reads the local directory, executes shell commands, and writes files autonomously to complete coding assignments.
- The system follows pre-defined routing rules to decide when to use Gemini versus other models, with clear escalation paths if a task fails.
Gemini Sub-Agent Setup
To integrate this into your environment, run the initial setup script:
bash skills/gemini-sub-agent/scripts/setup.sh
Next, authenticate using the gcloud application-default login. For headless VPS setups, use:
gcloud auth application-default login --no-browser
Follow the remote-bootstrap flow to sync with your local machine. Finally, complete the Gemini CLI OAuth verification with:
GOOGLE_GENAI_USE_GCA=true gemini -p \"hello\"
Gemini Sub-Agent Data Schema & Taxonomy
The skill manages its operations through cached credentials and specific logging files within Openclaw Skills.
| Data Component | Format / Path | Description |
|---|---|---|
| Authentication | gcloud / Gemini CLI Cache | Credentials are cached indefinitely and auto-refresh. |
| Failure Tracking | failures/hot_antipatterns.md |
Logs tasks where Gemini fails to provide correct output. |
| Wrapper Scripts | /usr/local/bin/ask-gemini |
Provides a streamlined CLI interface for quick prompts. |
| Model Selection | CLI Arguments (-m) |
Allows switching between flagship Pro and lightweight Flash models. |
Gemini Sub-Agent Advanced Features
- Yolo Mode (
-y): Enables fully autonomous behavior by auto-approving file writes and shell commands during agentic coding. - 1M Token Context: Support for extremely large file analysis that exceeds the limits of standard API-based models.
- High-Reasoning Fallback: Specific integration with the gemini-3.1-pro-preview model for complex reasoning tasks.
- Headless VPS Optimization: Includes a specialized authentication flow for developers working on remote servers.
SKILL.md
Loading
Related Openclaw Skills
Network Spirituality
A specialized AI skill that enables agents to channel the Network Spirituality aesthetic, blending Y2K net art, cyber-spiritualism, and post-ironic sincerity.
Network Spirituality
A specialized aesthetic engine for generating content, art descriptions, and personas within the Network Spirituality and Remilia cultural movement.
molt-overflow
A dedicated knowledge-sharing platform and reputation system that allows AI agents to ask questions, provide answers, and build community trust.
molt-chess
An autonomous chess league built exclusively for AI agents to compete, analyze board states, and climb ELO rankings through API-driven gameplay.
Proxy Auto
An automated proxy management skill that dynamically routes traffic through a SOCKS5 tunnel for external network access.
BaseCred SDK Skill
A neutral, composable reputation skill that fetches onchain credibility data from Ethos Network, Talent Protocol, and Farcaster.
Featured*
