repo-agents-pack

name	repo-agents-pack
description	Generate AGENTS.md — agent configuration documentation for any repository. Detects skills, MCP servers, and workflows from repo structure, then produces an actionable agent configuration file that teaches AI agents how to work with the codebase. Use when users want to create agent-compatible documentation, configure Copilot for a repo, or set up MCP server integration.
compatibility	Works with any Git repository. Pairs with the repo-onboarding-pack skill. Uses Microsoft Learn MCP Server (https://learn.microsoft.com/api/mcp) for Microsoft technology verification. Integrates with GitHub Copilot SDK (@github/copilot-sdk).

Repo Agents Pack Generator

Create AGENTS.md — a structured agent configuration document that teaches AI agents (GitHub Copilot, custom agents) how to work effectively with a repository.

Trigger Phrases

This skill activates when users say things like:

• "Generate agents config for this repo"
• "Create AGENTS.md for this project"
• "Configure Copilot skills for this repo"
• "What MCP servers should this repo use?"
• "Set up agent workflows for this codebase"
• "Create agent onboarding docs"

What This Skill Generates

Document	Purpose
AGENTS.md	Skills, MCP servers, and workflows for AI agents working with this repo

Relationship to Other Skills

This skill is invoked automatically by repo-onboarding-pack as part of the 9-step pipeline (AGENTS.md is generated in step 7: Compile). It can also be used standalone when users only need agent configuration.

Companion Skill	How It Helps
repo-onboarding-pack	Provides repo metadata (languages, deps, structure) that feeds AGENTS.md generation
microsoft-docs	Validates Microsoft technology references in detected skills
microsoft-code-reference	Verifies SDK/API patterns referenced in agent skills
microsoft-skill-creator	Creates new dedicated skills for technologies detected during scanning

AGENTS.md Structure

Required Sections

Section	Content	Minimum
Header	Project name and one-line description	Must name actual project
Skills	Table of agent skills with triggers	At least 1 skill per detected language
MCP Servers	Configured MCP servers with tool lists	microsoft-learn always included
Workflows	Step-by-step workflows agents should follow	At least 1 (onboarding workflow) + code-review
How to Use with GitHub Copilot	Example prompts for using Copilot with the repo	5+ example prompts

Output Template

# {projectName} — Agent Configuration

{One-sentence description of the project and its tech stack.}

## Skills

| Skill | Description | Triggers |
|-------|-------------|----------|
| {name} | {what it does} | {comma-separated trigger phrases} |

## MCP Servers

### {server-name}

**URL:** `{server-url}`

**Tools:** `{tool1}`, `{tool2}`, `{tool3}`

## Workflows

### {workflow-name}

{Description}

1. {Step 1}
2. {Step 2}
3. {Step 3}

Workflow

Step 1: Detect Skills from Repository

Analyze the repository to identify agent skills. Each skill represents a capability an AI agent should have when working with this codebase.

Build System Skills

Build File	Skill Name	Triggers
package.json (npm)	npm-build	build, install dependencies, compile
*.csproj (.NET)	dotnet-build	build, restore packages, compile
Cargo.toml (Rust)	cargo-build	build, compile, check
go.mod (Go)	go-build	build, compile, test
requirements.txt / pyproject.toml	python-build	install deps, build, package
Makefile	make-build	build, compile, clean

Test Framework Skills

Detection	Skill Name	Triggers
Jest / Vitest / Mocha	test-runner	run tests, test coverage, check tests
pytest / unittest	python-test	run tests, pytest, coverage
xUnit / NUnit / MSTest	dotnet-test	run tests, dotnet test
Go test	go-test	run tests, go test

Language-Specific Skills

For each detected language (top 3 by file count), generate a development skill:

{language-lowercase}-development
Description: Develop and review {Language} code
Triggers: write {Language}, review {Language}, refactor {Language}

Microsoft Technology Skills

When Microsoft technologies are detected, add specialized skills:

Detection	Skill	Triggers
@azure/* packages	azure-services	deploy, configure Azure, manage resources
.bicep files	infrastructure	deploy infrastructure, Bicep, IaC
host.json (Functions)	azure-functions	deploy function, add trigger, test function
Microsoft.SemanticKernel	semantic-kernel	AI plugins, kernel, agents
@microsoft/microsoft-graph	graph-api	Graph queries, user data, Teams

Step 2: Detect MCP Servers

Identify MCP servers the repository should connect to.

Always Include

### microsoft-learn

**URL:** `https://learn.microsoft.com/api/mcp`

**Tools:** `microsoft_docs_search`, `microsoft_docs_fetch`, `microsoft_code_sample_search`

Conditional MCP Servers

Detection	MCP Server	Tools
.mcp.json exists in repo	Parse and include all configured servers	As configured
Azure dependencies detected	Suggest Azure MCP if not already configured	Azure resource tools
GitHub Actions detected	Suggest GitHub MCP	PR, issue, workflow tools

Step 3: Detect Workflows

Generate step-by-step workflows that agents should follow for common tasks.

Required Workflow: Onboarding

Always generate this workflow:

### onboarding

New contributor onboarding workflow

1. Clone repository
2. Install dependencies
3. Run tests
4. Read ONBOARDING.md

Conditional Workflows

Detection	Workflow	Steps
Always included	onboarding	Clone → Install → Test → Read docs
Always included	code-review	Read PR → Review diff → Check style → Verify tests → Run tests → Give feedback
Build scripts present	development	Create branch → Make changes → Run tests → Build → Submit PR
CI config detected	ci-cd	Push to branch → CI tests → CI build → Deploy on merge
Docker/container files	containerization	Build image → Run locally → Test → Push to registry
.bicep / azuredeploy.json	infrastructure	Edit Bicep → Validate → What-if → Deploy
Notebook files (.ipynb)	experimentation	Create notebook → Run cells → Review output → Export results

Step 4: Quality Validation

After generating AGENTS.md, verify:

Check	Pass Criteria
Skills count	At least 1 per detected language + 1 for build system
Skill triggers	Each skill has 2+ trigger phrases
MCP servers	microsoft-learn is always present
Workflows	At least 1 workflow (onboarding)
File references	No fabricated file paths
Technology names	Match actual detected dependencies

Quality Checklist

Before completing, verify:

• AGENTS.md header names the actual project
• Every detected language has a corresponding skill
• Build system skill matches the actual build tool (npm, dotnet, cargo, etc.)
• Test runner skill matches detected test frameworks
• microsoft-learn MCP server is included
• Any .mcp.json in the repo is parsed and its servers are included
• Onboarding workflow is present
• Code-review workflow is present
• Additional workflows match repo capabilities (CI, Docker, infra)
• "How to Use with GitHub Copilot" section with example prompts
• No hallucinated technologies or tools
• Skill trigger phrases are natural language and actionable

Example Output

For a TypeScript + Azure Functions repository:

# my-azure-functions — Agent Configuration

Agent configuration for my-azure-functions — a TypeScript project with 24 dependencies.

## Skills

| Skill | Description | Triggers |
|-------|-------------|----------|
| npm-build | Build and manage the npm project | build, install dependencies, compile |
| test-runner | Run tests using jest | run tests, test coverage, check tests |
| typescript-development | Develop and review TypeScript code | write TypeScript, review TypeScript, refactor TypeScript |
| azure-functions | Deploy and manage Azure Functions | deploy function, add trigger, test function |

## MCP Servers

### microsoft-learn

**URL:** `https://learn.microsoft.com/api/mcp`

**Tools:** `microsoft_docs_search`, `microsoft_docs_fetch`, `microsoft_code_sample_search`

## Workflows

### onboarding

New contributor onboarding workflow

1. Clone repository
2. Install dependencies
3. Run tests
4. Read ONBOARDING.md

### development

Standard development workflow

1. Create feature branch
2. Make changes
3. Run tests
4. Build project
5. Submit PR

### ci-cd

Continuous integration and deployment

1. Push to branch
2. CI runs tests
3. CI runs build
4. Deploy on merge to main

### code-review

Structured code review workflow for learning and quality assurance

1. Open the pull request and read the description
2. Review the diff file-by-file, starting with tests
3. Check code style and naming conventions
4. Verify tests cover the changes
5. Run the test suite locally
6. Leave constructive feedback with specific suggestions

## How to Use with GitHub Copilot

This repository is configured for GitHub Copilot Agent Mode. Use these example prompts in VS Code:

| Goal | Example Prompt |
|------|---------------|
| Understand the project | "Explain the architecture of this project" |
| Start a task | "Help me work on Task 1 from TASKS.md" |
| Review code | "Review the changes in my current branch" |
| Find documentation | "What does the API do?" |
| Debug an issue | "Why is this test failing?" |
| Learn a concept | "Explain how error handling works in this codebase" |

> **Tip:** Open the onboarding docs (ONBOARDING.md, RUNBOOK.md, TASKS.md) as context when chatting with Copilot for better answers.

Agent Invocation

User: "Generate agents config for this repo"

Agent:

1. Scan repository (or reuse metadata from repo-onboarding-pack)
2. Detect skills from languages, build system, test frameworks
3. Detect MCP servers from .mcp.json and technology analysis
4. Generate workflows from build scripts and CI config
5. Render AGENTS.md with quality validation
6. If Microsoft technologies detected, add specialized skills

References

• repo-onboarding-pack — Parent skill that invokes AGENTS.md generation
• microsoft-skill-creator — Create new skills for detected technologies
• microsoft-tech-verification.md — MCP validation guidance