Vibeleaderboard
Index / app
Visit github.com
Category
Developer Tools
Rank

Previous survey · No. 805 ·

Type
APP
Builder
dotnet
Date

About

Agent Skills that help AI coding agents work with .NET and C#.

What it does

A modular marketplace of specialist guidance for coding assistants. Install a focused plugin for work such as build diagnosis, data access, testing, upgrades, web development, diagnostics, package management, MAUI, or AI integration. The core plugin also connects agents to a C# language server through LSP.

Why it's ranked here

The breadth is unusually practical, spanning everyday development and narrower jobs such as framework migration and incident analysis. More importantly, the repository includes explicit evaluation infrastructure that compares skilled runs with baselines, tracks activation and timeouts, and accounts for token cost. The experimental notes also document regressions and remove guidance that adds noise.

What's good

You can install only the capability needed instead of loading one large body of guidance. Support covers several agent hosts and follows an open skill standard. The maintainers test whether skills improve outcomes, record cases where baseline models already perform well, and simplify skills when rigid workflows or verbose catalogs reduce quality.

Tradeoffs

VS Code plugin support is explicitly a preview and may require manual enablement. Some documented experiments show activation failures, timeouts, token overhead, and cases where added guidance degraded results. Several tasks already reach baseline quality ceilings, so installing more guidance does not automatically improve an agent. Host setup and update procedures also vary.

How to use it well

Best for teams doing sustained .NET work through Copilot CLI, Claude Code, Cursor, Codex CLI, or preview VS Code plugins. Add the narrow plugin for the current task, then update it on demand. It earns particular value in diagnostics, build failures, migrations, package work, and specialized testing. It does not replace the coding host, .NET toolchain, or human validation.

Technical notes+

eng/vally-adapter/adapt.mjs splits experiment output by eval file, invokes Vally comparisons, and gates results with an exact one-sided sign test plus a minimum trial floor. eng/vally-adapter/adapt.test.mjs exercises retries, comparison failures, underpowered evaluations, sign-test behavior, and verdict stability. eng/vally-adapter/consolidate.mjs builds Markdown summaries from per-skill results. eng/dashboard/dashboard.js renders quality and efficiency data, while eng/dashboard/token-usage.js aggregates agent and judge token usage. docs/agentic-workflows.md and docs/design/pr-triage-workflows.md describe repository automation and PR state handling.

Observed

Packaging
Repository marketplace containing 15 listed .NET plugins, with individual skill installation also supported.
Install surfaces
Plugin marketplace installation is documented for Copilot CLI, Claude Code, Cursor, and Codex CLI.
Editor support
VS Code and VS Code Insiders plugin support is documented as a preview feature.
Standard
Skills follow the agentskills.io open standard.
Interfaces
The core dotnet plugin includes C# language server integration through LSP.
Evaluation structure
Repository includes comparison adapters, automated tests, result consolidation, dashboards, and experimental design notes.

Read from README.md, docs/agentic-workflows.md, docs/design/pr-triage-workflows.md, docs/dotnet-experimental/exp-mock-usage-analysis-design-notes.md, docs/dotnet-experimental/exp-test-maintainability-design-notes.md, eng/dashboard/dashboard.js, eng/vally-adapter/adapt.mjs, eng/dashboard/token-usage.js, eng/vally-adapter/adapt.test.mjs, eng/vally-adapter/consolidate.mjs.

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.