ATP Task Classification Ontology
Version: 0.2.0 | Canonical URI: https://agenttrustprotocol.org/ontology/v0.2.0
This page is the canonical, versioned home of the ATP task classification ontology referenced by the ontology_uri field in registrations and Proof metadata. It combines the U.S. Department of Labor O*NET occupational taxonomy and the HuggingFace task capability taxonomy, extended with ATP-specific categories. See Specification §6 for how classification is used in registration and Proofs.
Classification Fields
| Field | Source | Format | Example |
|---|---|---|---|
occupation | U.S. DOL O*NET SOC | 8-digit SOC code | "29-1141.00" (Registered Nurses) |
work_activities | O*NET Work Activities Taxonomy | Array of activity identifiers | ["4.A.1.b.2.I03.D03"] (Analyze test results) |
capabilities | HuggingFace Task Taxonomy + ATP Extensions | Array of capability identifiers | ["document-question-answering"] |
description | Free-form | Natural language text | Any task-specific details |
Capability Registry
This registry is exhaustive for ontology version 0.2.0; additional identifiers MUST NOT be used unless registered as ATP extensions via the ontology update process.
Code: code-completion, code-debugging, code-explanation, code-generation, code-translation, reinforcement-learning, test-generation
Multimodal: document-question-answering, image-text-to-text, table-question-answering, video-text-to-text, visual-question-answering
NLP: audio-classification, automatic-speech-recognition, conversational, feature-extraction, fill-mask, question-answering, sentence-similarity, summarization, tabular-classification, tabular-regression, text2text-generation, text-classification, text-generation, text-to-speech, time-series-forecasting, token-classification, translation, zero-shot-classification
Systems (ATP Extensions): configuration-management, incident-response, infrastructure-automation, log-analysis, performance-optimization, robotics, security-scanning
Vision: depth-estimation, image-classification, image-segmentation, image-to-image, image-to-text, object-detection, text-to-image, unconditional-image-generation, video-classification, video-generation, zero-shot-image-classification
Versioning
Ontology versions are released alongside specification versions but may also revise independently through the extension registration process. Prior versions remain resolvable at their versioned URIs.