HVE Core plugin
A Microsoft plugin for GitHub Copilot: an agent drives a task through the RPI phases while instructions pull in coding and security standards.
- 1,613 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Implementation agents edit project files, and the backlog agent changes work items in Azure DevOps, GitHub and Jira through MCP tools.
- Automatic mode runs the phases without routine confirmation prompts once you agree to it.
Install
- Works in
Manual install
copilot plugin marketplace add microsoft/hve-coreThe first README command for Copilot CLI. Then `copilot plugin install hve-core@hve-core`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Implementation agents edit project files, and the backlog agent changes work items in Azure DevOps, GitHub and Jira through MCP tools.
- Automatic mode runs the phases without routine confirmation prompts once you agree to it.
This is third-party code. Review the repository files before installing.
What it does
HVE Core builds Microsoft agents, instructions and skills for GitHub Copilot around a four-phase loop: research, plan, implement, review. The «RPI Agent» takes a task through the phases, and each phase can also be started alone with `/rpi-research`, `/rpi-plan`, `/rpi-implement` and `/rpi-review`. In manual mode the agent stays in the current phase until you say otherwise, and automatic mode starts only after explicit confirmation. Instructions apply coding standards on their own, while separate agents run documents and checks: ADR, BRD and PRD, a STRIDE threat model, an OWASP vulnerability review, architecture review and backlog work in Azure DevOps, GitHub and Jira. Artifacts of each phase are written to `.copilot-tracking/` under a date and task name.
Who it is for. Teams on GitHub Copilot that want one agent workflow with shared standards and checks instead of personal prompts.
Good fit when
- A large task needs code research first, then a plan, and only then edits.
- The team introduces shared coding standards and the agent should apply them unprompted.
- A release needs an OWASP vulnerability review or a STRIDE threat model.
- Requirements are easier to write as a PRD or ADR in a dialogue with an agent than from scratch in an editor.
- The backlog lives in Azure DevOps, GitHub or Jira and has to be worked through with an agent.
Not a fit when
- You need a stable production foundation: the authors call the project fast-moving and not backward compatible.
- The team has no GitHub Copilot subscription.
- A two-line fix: four phases with artifacts only slow the work down.
Example request
Research, plan, implement and review adding SHA validation for Docker images, starting with research.Limitations
The authors warn that the project is highly opinionated and changes quickly: workflows and interfaces may change without backward compatibility, so take patterns from it rather than building a product on it. A GitHub Copilot subscription and either VS Code or Copilot CLI are required. Canvases such as RPI Workflow work only through the plugin, since the VS Code extension does not carry them. Some agents are marked experimental. The meeting analysis agent needs the work-iq-mcp agreement, and transcripts may contain personal data and sit unencrypted on disk.
How to disable. Remove the plugin in Copilot CLI or the HVE Core extension in VS Code. Clear the `.copilot-tracking/` directory with phase artifacts by hand.
FAQ
How do I install HVE Core in GitHub Copilot?
In VS Code install the `ise-hve-essentials.hve-core` extension from the marketplace. For Copilot CLI run `copilot plugin marketplace add microsoft/hve-core`, then `copilot plugin install hve-core@hve-core`. Both routes carry the same set of components.
What does RPI stand for?
Research, Plan, Implement, followed by Review. Each phase has its own command, `/rpi-research`, `/rpi-plan`, `/rpi-implement` or `/rpi-review`, and the «RPI Agent» walks through all four.
Will the agent change code without my consent?
Not by default: manual mode stays in one phase until you give the next command. Automatic mode, Full Auto, starts only after explicit confirmation. Blockers, required human review, and destructive or externally visible actions still need your answer.
Where do the phase results go?
Into `.copilot-tracking/`: separate subfolders for research, plans, plan critiques, changes and review logs. Files are named by date and task name.
Can I take only part of the components?
Yes. The `hve-core-installer` skill copies chosen agents, instructions and skill directories into your repository and records the result in `.hve-tracking.json`. It does not copy hooks.
Discussion
No questions or reviews yet.
Related
PluginsRules, skills, agents and plugins for .NET: DDD, clean architecture, testing and LSP plugins for Copilot CLI and Claude Code.
Agents, instructions, skills and MCP servers for GitHub Copilot on the Nav stack and a nav-pilot CLI guiding tasks in phases
Codev runs a GitHub issue through the SPIR protocol: specification, plan, implementation and review, with parallel builders.
A Grid Dynamics plugin that loads a team's shared engineering standards, architecture and workflows into a coding agent session
Orchestrator skill: from an idea the agent writes requirements and a spec, splits work into tasks, builds the project, checks it
A skill that analyzes a repository and generates files for AI agents: AGENTS.md, pointers, reviewers, CI and templates
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| HVE Core (this card) | no | 1,613 | |
| Spec Kit | no | 149,633 | drives development from a specification through the specify CLI; HVE Core is built on the RPI loop and ships ready standards and checks |
| BMad Method | no | 55,338 | carries a project from idea through roles and stories; HVE Core runs inside GitHub Copilot with research, plan, implement and review phases |
README in short
The project rests on the idea that AI carries rules and context while people keep the decisions. To start, install the extension or plugin and pick the «RPI Agent» in Copilot Chat. The rest of the documentation is split by role: getting started, team adoption and contributing.