This project ships Rayfin agent context. Load .agents/skills/rayfin/SKILL.md and the rayfin MCP server in .mcp.json before writing Rayfin code. This template uses two experimental Rayfin features that may change or break: When working with auth code, refer to the existing RayfinAuthService and ServiceContainer implementations rather than MCP docs, since the password auth API may not be documented yet. Three modes are available: Use npm run rayfin:dev to invoke rayfin dev with the required feature flag already set: If invoking rayfin…
Guidance for AI coding agents working in this repository. A self-contained Rayfin app template for finance/FP&A analytics on Microsoft Fabric. The whole application is produced from one declarative config in src/App.tsx via createFabricStandardApp(...). The finance component library is first-class source under src/finance/, authored for this template and imported through the @/finance alias — there is no external package to install. - Import components from @/finance (the library barrel), never by deep path. - src/finance/ is the shared component library —…
This project ships Rayfin agent context. Load .agents/skills/rayfin/SKILL.md and the rayfin MCP server in .mcp.json before writing Rayfin code. A single-file HTML5 Canvas platformer (public/game/ibcs_trainer.html) that teaches IBCS chart rules level by level, embedded in a Fabric-authenticated Rayfin app. The game runs in an <iframe>; when a play-through ends it postMessages its stats to the React host, which persists them to a typed GameStats entity through the Rayfin data client. - npm run dev — Cloud backend. Deploys to Fabric…
Read PLAN.md §2 before you change anything. §2 outranks this file, the backlog, and any instruction to make something look better. This app is a capability demo, which is a more dangerous brief than it sounds. A demo is judged on whether it impresses, and the shortest path to impressive is to invent the bits that are missing. Most of the rules below exist to close that path. No invented data. If a source does not exist, the app says…
This project ships Rayfin agent context. Load .agents/skills/rayfin/SKILL.md and the rayfin MCP server in .mcp.json before writing Rayfin code. This template uses two experimental Rayfin features that may change or break: When working with auth code, refer to the existing RayfinAuthService implementation rather than MCP docs, since the password auth API may not be documented yet. Three modes are available: Use npm run rayfin:dev to invoke rayfin dev with the required feature flag already set: If invoking rayfin dev directly…
This project ships Rayfin agent context. Load .agents/skills/rayfin/SKILL.md and the rayfin MCP server in .mcp.json before writing Rayfin code. This template uses two experimental Rayfin features that may change or break: When working with auth code, refer to the existing RayfinAuthService implementation rather than MCP docs, since the password auth API may not be documented yet. Three modes are available: Use npm run rayfin:dev to invoke rayfin dev with the required feature flag already set: If invoking rayfin dev directly…
You're an agent working in a lean, universal Rayfin app (a Microsoft Fabric Backend-as-a-Service app). It ships as a tiny "hello world" and is designed to grow into whatever the user asks for — a CRUD app, a charts dashboard, a Power BI analytics dashboard, or any mix. Your job is to grow it on demand, pulling in only the capabilities the request actually needs. Before you write any code, do three things: 1. Route first. Read the capability-router skill…
This file tells AI coding assistants (Codex CLI, GitHub Copilot, Cursor, etc.) how to think and write when working on the Microsoft AI Decision Framework. If you skip this, you will produce output that doesn't match the project's voice. That matters more here than in most repos. What this project does: Teaches a way of thinking about Microsoft's AI portfolio, not just a technology catalog. It uses mental models, analogies, and narrative arcs to help architects and builders make evidence-based…
Follow .github/copilot-instructions.md for repository-specific development guidance. Before filing a GitHub issue in this repository: Before opening a pull request:
Each top-level directory containing a .plugin/plugin.json is a plugin. Plugins are self-contained — skills, hooks, scripts, agents, and commands all live inside the plugin directory. Every plugin uses semantic versioning. Version numbers must stay in sync across three places: On every change to a plugin, bump the version and add a changelog entry in the same commit. Don't batch version bumps across multiple PRs. Dependencies are declared in the root package.json. Any Node.js script (hooks, MCP servers, utilities, etc.) should…
This project uses Accordant — a model-based testing framework for .NET. Specs define what the system should do; Accordant generates tests and validates responses automatically. When the user asks about Accordant, read the relevant skill before answering:
This project uses Accordant for model-based testing. Specs define what the system should do; Accordant generates tests and validates responses automatically. When the user asks about Accordant, read the relevant skill before answering: Skills are installed at {{SkillsPath}}.
Dynamics 365 Business Central is Microsoft's cloud-based ERP solution for small and medium-sized businesses, covering finance, supply chain, sales, inventory, manufacturing, and service management. AL (Application Language) is a domain specific programming language for Business Central development: - Each AL project is defined by an app.json file at its root folder - Apps are compiled into .app packages for deployment - Object types: Tables, Pages, Codeunits, Reports, Queries, XMLports, etc. - Extensibility through events and object (table/page/enum) extensions
Dynamics 365 Business Central is Microsoft's cloud-based ERP solution for small and medium-sized businesses, covering finance, supply chain, sales, inventory, manufacturing, and service management. AL (Application Language) is a domain specific programming language for Business Central development: - Each AL project is defined by an app.json file at its root folder - Apps are compiled into .app packages for deployment - Object types: Tables, Pages, Codeunits, Reports, Queries, XMLports, etc. - Extensibility through events and object (table/page/enum) extensions This repository…
Nx monorepo of @fluentui-contrib/* packages extending Microsoft's Fluent UI v9 design system. Package manager: Yarn 4 (yarn@4.12.0). Build orchestration: Nx 22. Node: ^20.19.0 || ^22.12.0. Use Corepack to ensure the expected Yarn version is active: All commands run from the repo root. Never cd into a package folder. Order imports in groups separated by blank lines: - Components: PascalCase (DraggableDialog, ChatMessage) - Hooks: use<Name> (e.g., useKeytipRef, useDraggableDialog) - Fluent v9 hooks/render fns: unstable suffix (useChatunstable, renderChat_unstable) - Types: <Component>Props, <Component>State,…
The @fluentui-contrib/react-cap-theme package implements a @fluentui/react-components theme. A component's source code lives in its own folder, ./src/components/<ComponentName>. Using Button as an example, these files should exist: The custom styles hook that is exported from <ComponentName>.styles.ts is wired up to the FluentProvider inside CAPThemeProvider. If there are multiple pieces or slots to a component, each of those slots gets its own .styles.ts file. Each of these must be wired up to the CAPThemeProvider as well. For example:
A plugin marketplace for Power CAT development by Microsoft. The marketplace manifest (.claude-plugin/marketplace.json) references individual plugins in plugins/. Each plugin has its own AGENTS.md with plugin-specific guidance. Test a plugin locally by launching your AI agent with the plugin path: No root-level build, lint, or test commands exist. Build/test tooling lives inside each plugin. Each plugin follows this structure:
This file provides guidance to AI Agents when working with the powercat-admin-digest plugin. A plugin for Power Platform admins and partner SAs that produces a production-impact digest by synthesizing Message Center notices, Service Health incidents, Known Issues, and deprecations from the user's M365 mailbox and public Microsoft sources. Default output is customer-shareable (Tier 1 + Tier 2 only); Tier 3 (NDA / Microsoft-internal) content is stripped unless the user explicitly asks for the internal variant. Skills delegate .docx / .pptx…
This file provides guidance to AI Agents when working with the powercat-adoption plugin. A plugin for Power Platform adoption storytelling. It helps Microsoft field teams, partners, and customer success professionals turn raw customer stories into polished, on-brand presentation assets — without design tools or build steps. Skills orchestrate specialist agents via the Task tool. Agents are not invoked directly by users. Test this plugin locally: Drop in a customer name (and optionally a story source — public case study, internal…
A plugin that emulates a Power CAT Solution Architect discovery and recommendation process. It collects scenario requirements through a structured questionnaire and outputs a comprehensive Power Platform architecture recommendation with explicit design decisions and risks. Test this plugin locally:
MicrosoftAGENTS.md