---
title: "hu-core"
description: "Core HU Agent Framework skills and project guardrails."
canonical: https://agentpluginsdirectory.com/plugins/hu-core
last-updated: 2026-09-25
---

# hu-core
Core HU Agent Framework skills and project guardrails.
- Slug: hu-core
- Publisher: Data Science Pool
- Repository: https://github.com/uashogeschoolutrecht/hu-agent-framework
- Manifest: plugins/hu-core/plugin.json
- Version: 0.2.0
- License: MIT
- Category (editorial): agent-tooling
- Skills: 11 (align-deep, align-quick, coordinate-tasks, data-model-policy, get-started, hu-knowledge, make-agent, make-skill, plan-tasks, refine-principles, share-contribution)
- MCP servers: 0
- Stars: 1
- Repository created: 2026-08-10
- Repository last pushed: 2026-08-20
- Publisher type: Organization
- Listing: https://agentpluginsdirectory.com/plugins/hu-core
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## Skills

- align-deep: Reduce material uncertainty before specification when a HU project is big, risky, ambiguous, or likely to cause substantial rework. Invoked by get-started after quick alignment; not a standalone entry point.
- align-quick: Establish the minimum shared understanding needed to start a HU project safely without creating unnecessary questioning friction. Invoked by get-started; not a standalone entry point.
- coordinate-tasks: Execute dependency-aware HU task files in human-approved, bounded waves using tightly scoped VS Code sub-agents. Use only when task files and an execution policy already exist; get-started invokes this when the policy enables sub-agents.
- data-model-policy: Explain HU rules for data, models, and external endpoints. Use when a project handles data or chooses an AI model or external service.
- get-started: Start or continue a HU project using a risk-aware, spec-driven development workflow. Use when beginning a project, adding a feature, or asking how to run project work with the HU framework. For authoring or sharing a skill or agent, use make-skill, make-agent, or share-contribution instead.
- hu-knowledge: Find and apply reviewed HU-wide knowledge before making development decisions, especially for HU terminology, organizational context, standards, templates, and common solution patterns.
- make-agent: Create a custom HU agent with a narrow role and an explicit tool allowlist. Use when a reusable role needs its own tool permissions or its own context, and a skill cannot express it.
- make-skill: Create a new HU skill, or improve an existing one, so it follows the framework's structure and triggers reliably. Use when the user wants to turn a repeated prompt, workflow, or piece of know-how into a reusable skill.
- plan-tasks: Turn an agreed HU specification into small, dependency-aware task files with local tests and intermediate vertical-slice validation tasks. Use when an agreed specification already exists; get-started invokes this at its decomposition stage.
- refine-principles: Capture and sharpen the constraints and preferences underneath a user's stated goal into project principles, and apply them to design and implementation choices. Use during deep alignment, when the user gives evaluative feedback on a decision or design, or when they ask what the project stands for.
- share-contribution: Check a skill or agent against the HU framework standard and prepare a reviewed pull request to share it with the team or university. Use when the user asks to share, contribute, publish, or promote a skill or an agent.

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.