HVE Core - Data Science

Evaluation dataset creation, data specification generation, Jupyter notebooks, and Streamlit dashboards

Generate data specifications, Jupyter notebooks, and Streamlit dashboards from natural language descriptions. Evaluate AI-powered data systems against Responsible AI standards. This collection includes specialized agents for data science workflows in Python and RAI assessment.

[!CAUTION] The RAI agents and prompts in this collection are assistive tools only. They do not replace qualified human review, organizational RAI review boards, or regulatory compliance programs. All AI-generated RAI artifacts must be reviewed and validated by qualified professionals before use. AI outputs may contain inaccuracies, miss critical risks, or produce recommendations that are incomplete or inappropriate for your context.

Included Artifacts

Chat Agents

Name	Description
eval-dataset-creator	Creates evaluation datasets and documentation for AI agent testing using interview-driven data curation
gen-data-spec	Generate comprehensive data dictionaries, machine-readable data profiles, and objective summaries for downstream analysis (EDA notebooks, dashboards) through guided discovery
gen-jupyter-notebook	Create structured exploratory data analysis Jupyter notebooks from available data sources and generated data dictionaries
gen-streamlit-dashboard	Develop a multi-page Streamlit dashboard
rai-planner	Responsible AI assessment planning agent with 6-phase conversational workflow. Guides planning against NIST AI RMF 1.0 as the default evaluation framework. Prepares RAI security model, impact assessment, control surface catalog, and dual-format backlog handoff.
researcher-subagent	Research subagent using search tools, read tools, fetch web page, github repo, and mcp tools
test-streamlit-dashboard	Automated testing for Streamlit dashboards using Playwright with issue tracking and reporting

Prompts

Name	Description
rai-capture	Initiate responsible AI assessment planning from existing knowledge using the RAI Planner agent in capture mode
rai-plan-from-prd	Initiate responsible AI assessment planning from PRD/BRD artifacts using the RAI Planner agent in from-prd mode
rai-plan-from-security-plan	Initiate responsible AI assessment planning from a completed Security Plan using the RAI Planner agent in from-security-plan mode (recommended)
synth-data-generate	Generate comprehensive synthetic data for any specified subject with realistic patterns and relationships

Instructions

Name	Description
coding-standards/python-script	Instructions for Python scripting implementation
coding-standards/uv-projects	Create and manage Python virtual environments using uv commands
rai-planning/rai-backlog-handoff	RAI review and backlog handoff for Phase 6: review rubric, RAI review summary, dual-format backlog generation
rai-planning/rai-capture-coaching	Exploration-first questioning techniques for RAI capture mode adapted from Design Thinking research methods
rai-planning/rai-identity	RAI Planner identity, 6-phase orchestration, state management, and session recovery
rai-planning/rai-impact-assessment	RAI impact assessment for Phase 5: control surface taxonomy, evidence register, tradeoff documentation, and work item generation
rai-planning/rai-risk-classification	Risk classification screening for Phase 2: prohibited uses gate, risk indicator assessment, and depth tier assignment
rai-planning/rai-security-model	RAI security model analysis for Phase 4: AI STRIDE extensions, dual threat IDs, ML STRIDE matrix, and security model merge protocol
rai-planning/rai-standards	Embedded RAI standards for Phase 3: NIST AI RMF 1.0 trustworthiness characteristics, subcategory mappings, and framework isolation architecture
shared/hve-core-location	Important: hve-core is the repository containing this instruction file; Guidance: if a referenced prompt, instructions, agent, or script is missing in the current directory, fall back to this hve-core location by walking up this file's directory tree.

Getting Started

After installing this extension, the chat agents are available in GitHub Copilot Chat:

Use custom agents by selecting the custom agent from the agent picker drop-down list in Copilot Chat
Apply prompts through the Copilot Chat interface
Reference instructions: they are automatically applied based on file patterns

Post-Installation Setup

Some chat agents create workflow artifacts in your project directory. See the installation guide for recommended .gitignore configuration and other setup details.

Browse All Collections

This extension is part of the HVE ecosystem.

See the full list of available collections and capabilities:

Browse Extension Collections

Pre-release Channel

HVE Core offers two installation channels:

Channel	Description	Maturity Levels
Stable	Production-ready artifacts only	`stable`
Pre-release	Early access to new features and experimental artifacts	`stable`, `preview`, `experimental`

To install the pre-release version, select Install Pre-Release Version from the extension page in VS Code.

Full Edition

Looking for more agents covering additional domains? Check out the full HVE Core extension.

Requirements

VS Code version 1.106.1 or higher
GitHub Copilot extension

License

MIT License - see LICENSE for details

Support

For issues, questions, or contributions, visit the GitHub repository.

Brought to you by Microsoft ISE HVE Essentials