Schema-driven multi-axis text classification that turns arbitrary text into structured JSON labels with Openclaw Skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install classify-structured
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 classify-structured 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 Classify Structured?
Classify Structured is an Openclaw Skills tool for turning unstructured text into a machine-readable JSON classification across multiple axes in a single command. Instead of running separate prompts for sentiment, topic, intent, or any other label set, you provide a JSON schema that defines the allowed labels for each dimension and receive one chosen label per axis.
It is designed for reliable automation: the workflow uses ZeroGPU, strict schema constraints, and shell-safe quoting so even multi-line text and special characters can be passed safely. That makes it a strong fit for routing, moderation, analytics, triage, and any pipeline that needs structured labels from natural language with minimal manual cleanup.
Classify Structured Use Cases
- Classifying customer feedback by sentiment, topic, and urgency in one pass.
- Routing support tickets to the correct queue based on intent and category.
- Tagging product reviews, survey responses, or community posts with consistent labels.
- Building evaluation datasets where each example needs schema-bound annotations.
- Normalizing free-form text into a JSON object that downstream systems can consume.
- Using Openclaw Skills in automation pipelines where deterministic structure matters more than verbose analysis.
How Classify Structured Works
- Define a JSON schema that maps each classification axis to its allowed labels, such as
sentiment,topic, orintent. - Wrap the source text in a heredoc command substitution so arbitrary multi-line content, quotes, and special characters survive shell parsing.
- Run
zerogpu classify_structuredwith the text argument and the-sschema flag. - The ZeroGPU-backed classifier returns a JSON object with one selected label per axis, constrained to the labels you defined.
- If the command output literally includes a line starting with
💰 ZeroGPU savings, append that exact line unchanged as the last line of the final response; otherwise, omit any savings note. - Use the JSON output directly in dashboards, routing rules, enrichment jobs, or other Openclaw Skills-driven workflows.
Classify Structured Setup
- Ensure the ZeroGPU CLI is available in your environment as
zerogpuand that you can run Bash commands. - Prepare a schema that uses axis names as keys and arrays of allowed labels as values.
- Pass the source text verbatim through a heredoc so quoting never breaks on punctuation or newlines.
- Run a test classification with your preferred axes:
zerogpu classify_structured "$(cat <<'ZGPU_T'
Your source text goes here.
It can span multiple lines and include special characters.
ZGPU_T
)" -s '{"sentiment":["positive","negative","neutral"],"topic":["support","billing","product"]}'
- Treat the result as structured JSON and feed it into whatever system consumes your Openclaw Skills output.
- Only append the
💰 ZeroGPU savingsline if it appears literally in the command output; do not invent or infer it.
Classify Structured Data Schema & Taxonomy
Openclaw Skills organizes data entirely around a schema-driven classification contract. There are no persistent files generated by default; the primary artifacts are the input text, the schema, and the JSON result.
| Component | Type | Purpose |
|---|---|---|
| Source text | String / multiline string | The arbitrary text to classify. |
| Schema | Single-quoted JSON object | Maps each axis to its allowed labels. |
| Axis | JSON object key | A classification dimension such as sentiment or topic. |
| Allowed labels | Array of strings | The only valid outputs for each axis. |
| Result | JSON object | Returns one selected label per axis. |
| Savings note | Optional plain-text line | Only present if the output literally contains 💰 ZeroGPU savings. |
Taxonomy rules
- Each axis is explicit and schema-defined.
- Each axis may only return one label.
- Labels must come from the allowed set provided in the schema.
- The output is intended to remain clean JSON for downstream automation.
- The taxonomy is user-defined, so Openclaw Skills can adapt to sentiment, topic, intent, routing, moderation, or custom labeling schemes.
Classify Structured Advanced Features
- Multi-axis classification in a single ZeroGPU call, reducing pipeline complexity.
- Schema-enforced label selection that prevents free-form drift and keeps outputs predictable.
- Heredoc-safe quoting for arbitrary text, including multi-line content and special characters.
- JSON-first output that plugs directly into scripts, ETL jobs, dashboards, and agent workflows.
- Optional passthrough of the exact
💰 ZeroGPU savingsline when present in the CLI output. - A clean fit for Openclaw Skills automation where structured labels are more useful than verbose explanations.
SKILL.md
Loading
Related Openclaw Skills
redact-pii
Openclaw Skills for redact-pii masks sensitive text in-line with ZeroGPU, replacing detected PII with bracketed labels before you share or log it.
Social Arbitrage
Openclaw Skills turns real-world social and consumer shifts into evidence-backed U.S. equity research with a skeptical, Chris Camillo-inspired workflow.
World Builder
A meta-skill that turns a world idea or character set into a coherent, 12-dimensional world specification.
CertainLogic Stack
CertainLogic Stack is an Openclaw Skills AI stack that combines verified knowledge, low-cost routing, caching, and audit trails to cut spend and stop hallucinations.
Extract Entities
Openclaw Skills turns ZeroGPU into a fast, custom-label NER workflow for extracting entities from any passage of text.
extract-pii
extract-pii finds and groups personally identifiable information in raw text with ZeroGPU while leaving the source unchanged.
Featured*
