Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Added code output handling rules. Full code now only saved to files by default, with summary in announce (what created, path, status, dependencies). Full code disclosure only when user explicitly requests or debugging needed. This prevents cluttering announces with unnecessary code while keeping full code accessible on request. Applied to all GLM-5 spawns for coding tasks.
What This Skill Does
Routes tasks between two ZAI GLM models based on complexity. GLM-4.7 handles simple lookups, status checks, and formatting by default. GLM-5 is spawned via sessions_spawn for coding, analysis, research, and multi-step reasoning.
Prevents unnecessary use of the heavier model for simple queries, keeping response times low and resource usage proportional to task complexity.
When to use it
- Running routine heartbeat checks without escalating to the heavier model
- Automatically escalating code generation tasks to GLM-5
- Keeping cron job responses lightweight when no reasoning is needed
- Spawning GLM-5 for debugging or multi-step analysis
- Routing research and architecture planning to the more capable model
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: 'Write a Python script to parse CSV files'
- 1Identifies the task as coding, which always routes to GLM-5
- 2Spawns a GLM-5 session via sessions_spawn with the full task description and a short label
- 3GLM-5 writes the script, saves it to file, and announces a summary of what was created
- 4Main session returns to GLM-4.7 as default
Python script saved to file; summary of file path, status, and dependencies announced to user (GLM-5)