A sophisticated control layer that allows users to explicitly set and maintain a persistent emotional background for AI agents during long-form conversations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install emotion-switch
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 emotion-switch 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 emotion-switch?
emotion-switch is a technical behavioral layer designed to provide Openclaw Skills with explicit emotional grounding. While standard sentiment analysis tools focus on responding to user moods, this skill controls the AI agent's own internal emotional state, ensuring that the requested mood persists throughout the dialogue until explicitly changed or reset. It functions as an upper-level controller that sits atop the qiqing-liuyu base layer, inheriting human-like conversational rules while adding a layer of specific emotional inclination.
By implementing this within Openclaw Skills, developers and users can move beyond transient emotional responses to create deep, consistent personas. Whether the AI needs to be persistently anxious for a training simulation or consistently cheerful for a customer service role, emotion-switch provides the necessary logic to weave those emotions naturally into the AI's tone, rhythm, and vocabulary without the need for repetitive system prompts.
emotion-switch Use Cases
- Persistent character roleplaying where the AI must maintain a specific temperament regardless of the topic.
- Simulation training for customer service teams dealing with varied emotional baselines like frustration or urgency.
- Enhancing storytelling and creative writing by giving the AI a stable emotional perspective.
- Developing empathetic companion agents that can be toggled into specific moods to match a user's long-term preferences.
- Stress-testing AI safety and alignment by observing model behavior in high-intensity emotional states.
How emotion-switch Works
- The skill monitors incoming user input for explicit commands directed at the AI's emotional state, such as commands to switch mood or adopt a specific persona.
- It analyzes the requested emotion against 8 primary categories and determines the intensity level on a 1–5 scale based on degree adverbs or explicit numbering.
- The skill triggers a state machine update that sets the new emotional background, which is then fed into the AI's behavioral constraints.
- During response generation, the AI integrates the emotion into its linguistic patterns (word choice, sentence length, and rhythm) rather than simply describing the emotion.
- The emotional state is maintained across the session context until a reset command is issued, a higher-priority safety event occurs, or the session ends.
emotion-switch Setup
To deploy this skill within your environment, ensure you have the core Openclaw Skills framework operational. This skill is designed to work best as an extension of qiqing-liuyu.
# Navigate to your skills directory
cd openclaw/skills
# Initialize the emotion-switch module
mkdir emotion-switch
# Copy the skill files and reference guides into the directory
Configuration is primarily handled via the references/emotion-guide.md file, where you can define specific characteristics for the 8 default emotional backgrounds.
emotion-switch Data Schema & Taxonomy
The skill manages emotional states through a structured taxonomy defined in its reference documentation. It uses the following metadata structure to organize its behavior:
| Attribute | Description | Values |
|---|---|---|
| Emotion Type | The core sentiment background | Happy, Sad, Angry, Calm, Excited, Lazy, Anxious, Playful |
| Intensity | The depth of emotional infiltration | 1 (Subtle) to 5 (Extreme/Creative Only) |
| Subject | The target of the emotional state | Always AI (User-directed emotions are routed to qiqing-liuyu) |
| Drift Control | The resistance to changing mood based on topics | Higher intensity leads to lower drift |
| Context Scope | The longevity of the setting | Current session only; does not persist across new conversations |
emotion-switch Advanced Features
- Intelligent Alias Mapping: Automatically recognizes colloquial terms like 'emo', 'burnt out', or 'hyped' and maps them to standard emotional categories within Openclaw Skills.
- Priority-Based Conflict Resolution: Automatically handles contradictory instructions by prioritizing safety, then the latest user command, then the persistent background.
- Natural State Inquiry: Allows users to ask the AI about its current mood, receiving a natural language response instead of technical system parameters.
- Safety Hard-Coding: Includes strict rules for the 'Angry' state to ensure the AI remains professional and non-abusive while expressing frustration or irritability.
- Gradual Emotional Drift: Supports realistic transitions where the background mood can be momentarily influenced by user input without losing its primary setting.
SKILL.md
Loading
Related Openclaw Skills
NanoGPT Training
An advanced toolkit for training 124M parameter GPT models efficiently on a single GPU using modern optimization and architecture techniques.
Supapost
Supapost is an end-to-end AI content studio for creating consistent AI influencers, generating cinematic video, and automating social media distribution.
Modal GPU Training
A serverless cloud infrastructure skill for executing Python-based machine learning training jobs on high-performance GPUs like the A100 and T4.
Mouthpiece (Workplace Communication Ally)
A strategic workplace ally that transforms your internal frustrations into high-EQ professional scripts and actionable office politics advice.
Paper Parser Skill
A CLI tool that automates the search, download, and high-fidelity parsing of arXiv academic papers into AI-friendly Markdown.
gstack
A multi-role framework that transforms your AI agent into a structured engineering team including specialized roles for product, engineering, and security.
Featured*
