A deterministic rule-based engine that classifies clarified user input into task categories while assessing complexity and risk for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install input-classification-v1
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 input-classification-v1 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 Input Classification System?
The Input Classification System is a core architectural component designed to provide structure and predictability to AI agent operations. By processing clarified user input through a rigorous 15-category framework, this skill ensures that every request is assigned exactly one primary task type—such as debugging, code generation, or data analysis—alongside measurable complexity and risk scores. This deterministic approach eliminates ambiguity, allowing for precise downstream routing and execution planning within the Openclaw Skills ecosystem.
This system acts as the bridge between raw user intent and actionable technical plans. It utilizes a multi-step tie-breaking logic involving keyword dominance and action verb analysis to maintain high classification accuracy. By enforcing strict boundary definitions and state transitions, the Input Classification System prevents misclassification and ensures that the agent only proceeds when confidence thresholds are met, making it an essential tool for high-reliability AI coding environments.
Input Classification System Use Cases
- Routing user requests to specialized sub-agents based on the primary task category.
- Assessing the execution complexity and time estimates of a task before decomposition.
- Identifying high-risk operations that require human oversight or specific safety protocols.
- Enhancing the audit trail by logging precise classification decisions for every interaction in Openclaw Skills.
- Determining if further user clarification is needed when confidence scores fall below threshold.
How Input Classification System Works
- Validates that the input has been fully processed by a clarification system and marked as clarified.
- Scans the input for keywords and action verbs associated with the 15 predefined task categories.
- Applies deterministic tie-breaking rules, prioritizing high-urgency categories like DEBUGGING or DEPLOYMENT if multiple matches occur.
- Calculates a complexity level and risk score based on measurable operation counts and impact scope.
- Assigns up to three secondary tags to provide additional context for the downstream execution engine.
- Generates a structured ClassificationResult and sets the state transition to proceed to task decomposition or escalation.
Input Classification System Setup
To integrate the Input Classification System into your environment, ensure you have the core agent framework installed. This skill requires the clarification state to be satisfied before activation.
# Install the skill package
openclaw install input-classification-system
# Verify installation within your Openclaw Skills directory
openclaw list --installed
Ensure your agent configuration points to the system-integration.md reference for proper state transition handling.
Input Classification System Data Schema & Taxonomy
The Input Classification System outputs a structured ClassificationResult object. The data is organized into the following primary components:
| Field | Type | Description |
|---|---|---|
| primary_category | String | One of the 15 fixed codes (e.g., CG, DB, RF) |
| complexity_level | Integer | Scale of 1-5 based on estimated operations |
| risk_level | String | Low, Medium, High, or Critical |
| confidence_score | Float | Probability score (0.0 to 1.0) |
| secondary_tags | Array | Up to 3 additional context tags |
| state_transition | String | The next logical state for the agent |
All decisions are logged in the classification audit log for debugging and performance monitoring within Openclaw Skills.
Input Classification System Advanced Features
- Deterministic Tie-Breaking: Multi-stage logic using keyword frequency and action verb mapping to ensure 100% single-category output.
- Safety Escalation Triggers: Automatically halts execution and requests human intervention for high-risk or low-confidence results.
- Secondary Context Tagging: Supports up to 3 non-primary tags to preserve nuanced user intent across complex requests in Openclaw Skills.
- Category Boundary Definitions: Hard-coded rules that differentiate between overlapping tasks like Research vs. Analysis or Configuration vs. Deployment.
SKILL.md
Loading
Related Openclaw Skills
DuckDuckGo Web Search Skill
A privacy-focused web search skill that enables AI agents to retrieve real-time data, news, and multimedia without an API key.
DuckDuckGo Web Search
A privacy-focused web search skill that enables AI agents to retrieve real-time information, news, and images without user tracking.
Omnium Hub CRM Skill
A specialized AI skill for managing CRM contacts, opportunities, and appointments via the Omnium Hub API.
Video Production
A high-performance pipeline for generating, reviewing, and assembling multi-scene A/B videos using the Veo 3 API and FFmpeg.
Lite SQLite
A high-performance, ultra-lightweight SQLite database management skill designed for efficient local data persistence in AI agents.
Snapbyte Digest API
A specialized skill for fetching and formatting user-scoped developer news digests from the Snapbyte API into high-signal markdown.
Featured*
