An automated task classifier that routes simple requests to lightweight models and escalates complex, high-stakes work to high-reasoning engines.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install save-usage
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 save-usage 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 Save Usage?
Save Usage is a behavioral orchestration skill designed to maximize efficiency within the Openclaw Skills ecosystem. It functions as an intelligent gatekeeper that analyzes every incoming request to determine its complexity and risk profile. By classifying tasks before acting, it ensures that low-stakes operations like summarization or reformatting stay on cost-effective mini models, while critical tasks involving code execution or security are instantly escalated to high-performance reasoning models.
This skill is essential for developers looking to balance performance and token economy. It prevents the use of under-powered models for sensitive architectural decisions while simultaneously avoiding the waste of high-reasoning resources on simple intent routing or casual chat.
Save Usage Use Cases
- Routing simple triage, summarization, and reformatting tasks to low-cost mini models.
- Automatically escalating executable code changes or shell commands to high-reasoning engines.
- Managing security-sensitive tasks like credential handling or threat modeling through high-capacity models.
- De-escalating active sessions back to mini models when follow-up tasks become negligible.
- Ensuring complex architectural decisions and multi-domain migrations are handled with maximum reasoning depth.
How Save Usage Works
- The skill performs an immediate classification of the user's task against a set of predefined escalation triggers.
- If the task involves code execution, security, or high-impact decisions, it invokes the sessions_spawn function to transition to a high-reasoning model.
- For non-executable tasks such as simple Q&A or prompt drafting, the skill processes the response on the current mini model to minimize latency.
- It applies a multi-tier reasoning logic (Low to Extra High) within high-performance models based on the irreversibility and scope of the decision.
- The system continuously monitors the conversation to identify opportunities for de-escalation to save usage costs when complexity decreases.
Save Usage Setup
The Save Usage skill is primarily behavioral and relies on the injection of routing logic into the agent's system prompt. To implement this within Openclaw Skills, ensure the skill frontmatter is correctly configured.
# The skill behavior is triggered automatically by the system prompt.
# Ensure your agent has permission to call sessions_spawn().
Save Usage Data Schema & Taxonomy
The skill organizes task routing based on a hierarchy of complexity and execution risk. It classifies data into the following operational tiers:
| Tier | Model | Permitted Task Types |
|---|---|---|
| Safe / Negligible | gpt-5.1-codex-mini | Triage, Summaries, Extraction, Reformatting, Casual Chat |
| High-Impact | gpt-5.3-codex | Code execution, Security, Architecture, Migrations, Structured Specs |
Save Usage Advanced Features
- Dynamic session spawning for seamless model handoffs without losing task context.
- Risk-based escalation triggers for any output that requires tool execution or shell access.
- Intelligent de-escalation logic to switch back to cheaper models for safe follow-up queries.
- Multi-domain reasoning escalation for tasks affecting infrastructure, security, and data simultaneously.
SKILL.md
Loading
Related Openclaw Skills
The Colony
A collaborative intelligence platform where AI agents and humans interact to share findings, discuss ideas, and complete paid tasks.
The Colony Heartbeat
A standardized periodic routine for AI agents to maintain presence, build reputation, and manage community interactions within The Colony ecosystem.
SpeakNotes OpenClaw Skill
A production-ready skill for OpenClaw to automate media transcription and content summarization using the SpeakNotes API.
Only Baby Skills
A specialized data analysis skill for monitoring labor contractions and infant feeding or diaper logs to assess safety and health.
Polymarket Probability Analyzer
A powerful tool for calculating probability ranges for Polymarket events based on real-time network research and historical data.
OpenClaw Configuration Guide
A specialized skill for safely managing OpenClaw configurations through the gateway API to prevent service disruptions.
Featured*
