zown-gemini-governor

What This Skill Does

The zown-gemini-governor is an advanced orchestration layer designed to stabilize the Gemini interaction lifecycle within OpenClaw. It addresses the primary bottleneck of AI engineering: Token Per Minute (TPM) saturation. By implementing the 'Atomic Pipeline', this skill enforces strict context hygiene and real-time rate limit monitoring. It transforms unpredictable model behavior into a predictable, modular workflow by ensuring that context 'bloat'—which often leads to hallucination and rate-limit errors—is neutralized through mandatory pruning and systematic cool-downs. The skill acts as an intelligent supervisor, ensuring that your agent maintains operational throughput without hitting the 429 barrier.

Installation

To integrate this governor into your agent, execute the following command in your terminal within the OpenClaw environment: clawhub install openclaw/skills/skills/gtovd/zown-gemini-governor

Use Cases

• Heavy Software Engineering: Ideal for complex multi-file code refactors where session history normally balloons and triggers rate limits.
• Long-Running Research Agents: Perfect for tasks requiring massive data synthesis where memory needs to be condensed into MEMORY.md iteratively.
• High-Velocity Automation: Use when your agent is interacting with multiple APIs simultaneously and needs to regulate request frequency to maintain a 'perfect run' status.

Example Prompts

1. "Apply the zown-gemini-governor protocol: Prune the current project context and summarize our progress into MEMORY.md before we begin the next module."
2. "I am planning a major codebase refactor. Initialize the 9-stage atomic pipeline and let me know when you are ready for the first isolated step."
3. "We are at 60% TPM capacity. Execute the cooldown protocol for 60 seconds and sync our memory state."

Tips & Limitations

• Strict Adherence: The 50% Rule is not a suggestion; it is a hard limit designed to prevent service interruptions. Ignoring this may result in 429 errors.
• Context Pruning: Always keep your active files below 500 tokens. If you do not perform manual maintenance, the governor will force a context compaction.
• Atomic Logic: You cannot combine planning and execution. The governor requires distinct turns for both to maintain tracking accuracy. This skill does not replace standard model logic but acts as an external regulatory buffer.