Microsoft’s prefab prompt library makes Copilot follow house rules

What it does

HVE Core ships as a VS Code extension and CLI plugin that injects curated agents, prompts, instructions, and skills into GitHub Copilot. The underlying catalog includes 49 agents, 102 instructions, 63 prompts, and 11 skills, distributed through curated bundles that range from a 41-artifact flagship RPI (Research → Plan → Implement) collection to a complete 221-artifact library. The premise is convention-driven automation: the artifacts attempt to align Copilot’s output with repository-specific standards rather than one-size-fits-all suggestions.

The interesting bit

The project treats prompt engineering like a software dependency. Instead of leaving every developer to hand-craft Copilot prompts, Microsoft packages validated, versioned artifacts—complete with CI validation pipelines and quality gates—so teams can install a “coding standards” layer the same way they install a linter.

Key highlights

• 49 specialized agents (e.g., rpi-agent, task-researcher, memory) and 102 instructions that the README claims are auto-applied to matching files
• 63 reusable prompts and 11 skills with cross-platform scripts for tasks like commit messages and security planning
• Three distribution flavors: a flagship collection, an “All” bundle with 221 artifacts, and a selective installer for teams that want only specific domains
• Built-in validation scripts for linting, security, and quality checks, plus automated label-management workflows in the repository itself
• Some skills include third-party OWASP content licensed under CC BY-SA 4.0, clearly marked in frontmatter

Caveats

• The project is fundamentally a curated content pack and distribution mechanism for Copilot artifacts; the underlying reasoning still comes from GitHub Copilot itself
• The README prominently documents internal repository label-management workflows, which sits awkwardly alongside user-facing documentation
• While the README claims instructions are “auto-applied to matching files,” it does not explain the matching logic or how conflicts between multiple instructions are resolved

Verdict

Teams already paying for GitHub Copilot and struggling with inconsistent, generic suggestions will get the most value here—provided they are willing to adopt Microsoft’s prefab workflows. If you prefer hand-rolling prompts or do not use Copilot, this is a very well-organized prompt library you will never open.