A sophisticated logic framework for dynamically adjusting LLM temperature settings to ensure task-appropriate balance between deterministic precision and creative expression.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install dynamic-temperature
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 dynamic-temperature 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 Dynamic Temperature?
The Dynamic Temperature skill is a governance framework designed to optimize the performance of Large Language Models within the ecosystem of Openclaw Skills. By mapping specific task categories to a standardized temperature scale—ranging from 0.0 for high-stakes, irreversible actions to 0.8 for expressive, creative writing—this skill prevents common AI failures such as hallucinations in technical tasks or overly robotic responses in interpersonal communication. It provides a reliable decision-making structure that allows developers to maintain strict control over agent variance, ensuring that every interaction is grounded in the appropriate context.
Dynamic Temperature Use Cases
- Executing irreversible actions like deleting database records or sending official emails with absolute deterministic precision.
- Generating creative content, jokes, or icebreakers where higher variance and natural language flow are required.
- Coordinating schedules and calendar events where accuracy and factual consistency are non-negotiable.
- Drafting daily briefings and summaries that require a balance of structured thinking and readable warmth.
How Dynamic Temperature Works
- The agent analyzes the primary intent of the task to determine if it involves irreversible actions, technical scheduling, analytical summaries, or creative drafting.
- The system maps the identified task against the Openclaw Skills temperature scale, selecting a value between 0.0 and 0.8.
- The chosen temperature is applied to the LLM configuration or passed as a parameter when spawning specialized subagents.
- If the task involves communication or data deletion, the skill triggers an override rule to force a 0.0 temperature for the confirmation step.
- The agent generates the output with the optimized variance level, ensuring the response matches the required persona and technical constraints.
Dynamic Temperature Setup
To integrate this logic into your agent environment, you can define temperature defaults in your model configuration or specify them directly within the frontmatter of your Openclaw Skills.
# Example configuration for a deterministic agent
# Set in your agents.defaults.models:
temperature: 0.2
For dynamic usage, ensure your orchestration layer passes the temperature parameter during subagent initialization based on the task classification rules defined in the SKILL.md documentation.
Dynamic Temperature Data Schema & Taxonomy
The skill utilizes a taxonomy of task types and corresponding temperature values to govern agent behavior. This metadata ensures consistency across different Openclaw Skills implementations.
| Task Classification | Temperature | Logic Applied |
|---|---|---|
| Irreversible Actions | 0.0 | Zero variance, deterministic |
| Scheduling & Commands | 0.2 | High precision, factual |
| Summaries & Analysis | 0.3 | Structured, grounded |
| General Communication | 0.5 | Balanced, natural |
| Creative & Expressive | 0.8 | High variance, imaginative |
These rules are typically embedded in the skill frontmatter to guide the model's self-selection process.
Dynamic Temperature Advanced Features
- Communication Override System: Automatically forces 0.0 temperature for confirmation signals and reaction emojis to prevent unexpected variations.
- Per-Skill Optimization: Tailored recommendations for specialized agents like billing monitors or git backup tools.
- Task-Specific Subagent Spawning: Capability to programmatically adjust parameters when handing off high-precision tasks to sub-processes.
- Deterministic Reaction Signals: Ensures that standard replies like 'sure thing' remain exact strings without creative drift.
SKILL.md
Loading
Related Openclaw Skills
Daily Mindful
A digital sanctuary that generates a personalized, interactive mindfulness experience featuring guided breathing exercises and bilingual meditative content.
Inspector
A powerful integration for real-time application performance monitoring and automated data management via the Membrane framework.
Daily Fitness
A smart fitness companion that generates equipment-free, 5-15 minute daily workout routines complete with interactive timers and visual guides.
Finto.io Financial Data Integration
A powerful financial data aggregation skill for managing transactions and records via the Membrane platform.
Daily History Skill
An automated AI agent skill that generates stunning, bilingual visual timelines of historical events occurring on the current date.
BTC Price Tracker
A professional Bitcoin price monitoring and alert system featuring real-time tracking across multiple currencies and automated Telegram notifications.
Featured*
