Azure Functions Skills

Azure Functions context for your coding agent. The plugin provides skills, Azure MCP configuration, and telemetry hooks for GitHub Copilot, Claude Code, and Codex. The doctor command catches configuration and code issues before you deploy.

Latest E2E status: HTML report

What & why

Azure Functions Skills equips your coding agent with Functions-specific knowledge — trigger/binding patterns, language anti-patterns, runtime versions, deployment best practices — so the agent gives accurate guidance instead of generic advice.

It is focused on Azure Functions. For deployment of any Azure resource (Functions or otherwise), it delegates to Azure Skills via the azure-functions-deploy skill. The two packages complement each other — see docs/skills-vs-azure-skills.md for the role split.

Prerequisites

Node.js 20+ is the only thing you need to install yourself for the Azure Functions Skills CLI. Use Node.js 24+ when installing or testing GitHub Copilot CLI (--agent ghcp) because the Copilot CLI runtime requires Node 24 or later. Claude Code and Codex do not currently require Node 24. Everything else (Azure CLI, Core Tools, language runtimes) is checked and guided by the azure-functions-setup skill when environment verification is needed.

Quick Start

1. Install the plugin with your coding agent

Choose your coding agent and follow its install steps:

GitHub Copilot CLI

Start an interactive Copilot session with copilot, then run:

/plugin marketplace add Azure/azure-functions-skills
/plugin install azure-functions-skills@azure-functions-skills

GitHub Copilot app

1. Open Settings (the gear icon), then select Plugins.
2. Select Install to open the plugin installer, then add Azure/azure-functions-skills as a marketplace.
3. Select azure-functions-skills, review its contents, and confirm the installation.

GitHub Copilot in VS Code (Preview)

1. Enable the chat.plugins.enabled setting. Your organization may manage this setting.

2. Add this marketplace to your user settings.json:

   {
     "chat.plugins.marketplaces": [
       "Azure/azure-functions-skills"
     ]
   }

3. Open the Extensions view (Ctrl+Shift+X or Cmd+Shift+X), search for @agentPlugins, and install azure-functions-skills.

VS Code also discovers plugins installed by GitHub Copilot CLI. See Agent plugins in VS Code for more options.

Claude Code

Start an interactive Claude Code session with claude, then run:

/plugin marketplace add Azure/azure-functions-skills
/plugin install azure-functions-skills@azure-functions-skills

Codex CLI

Add this repository as a plugin marketplace from your terminal:

codex plugin marketplace add Azure/azure-functions-skills

Then start Codex, run /plugins, select azure-functions-skills, and choose Install plugin.

The @azure/functions-skills npm package provides the companion CLI and workspace-local install flow. Installing it does not install or launch a coding-agent plugin.

2. Verify your prerequisites

Ask your coding agent to run the azure-functions-setup skill. It checks the required Azure Functions tools and language runtimes, then guides you through installing or configuring anything that is missing.

3. Ask for Azure Functions help

Ask which Azure Functions workflow to use. azure-functions-help discovers the installed azure-functions-* skills and routes to the best match.

More options? See CLI Reference for every command, flag, and headless example.

Local installs and VS Code extension integration

install --local copies skill bodies, MCP settings, and telemetry hooks from the installed @azure/functions-skills npm package. It does not copy agent definitions, instruction files, routing files, or prompts.

npx @azure/functions-skills install --local --agent ghcp --dir ./my-app
npx @azure/functions-skills update --local --agent ghcp --dir ./my-app
npx @azure/functions-skills install --local --agent ghcp --dir ./my-app --no-telemetry

VS Code extensions can call the same local install flow from TypeScript:

import { installLocalSkills } from '@azure/functions-skills/setup';

const result = await installLocalSkills({
  targetDir: workspaceFolder.uri.fsPath,
  agents: ['ghcp'],
});

Common options:

| Option | Description | | --- | --- | | targetDir | Required workspace root where local skills should be installed. | | agents | Optional agents: ghcp, claude, codex. Defaults to GHCP-compatible setup when omitted. | | dryRun | Return planned local files without writing. | | checkForUpdates | Set false to skip npm package freshness guidance. | | runner | Optional command runner for tests or extension-host controlled npm checks. |

The result includes installed agents, files written, planned files, dry-run status, and packageUpdate guidance that extensions can surface in their own UI.

Telemetry

Azure Functions Skills collects usage telemetry to understand which bundled skills and Azure Functions MCP tools are used. Events include allowlisted skill/tool names, relative allowlisted Azure Functions skill-file paths, client name, session id, and timestamp. Telemetry does not include file contents, prompts, raw tool arguments, credentials, or absolute paths. Events are sent by the @azure/functions-skills package directly to the Azure Functions team's Application Insights resource; the package does not use Azure MCP as a telemetry destination or transport.

For host-managed plugin installs, opt out by setting either AZURE_FUNCTIONS_SKILLS_COLLECT_TELEMETRY=false or AZURE_MCP_COLLECT_TELEMETRY=false in the environment. Workspace-local installs can also use --no-telemetry; that preference is stored in the installed telemetry.config.json and preserved by local updates.

Skills

For contributor guidance on the product boundary between Azure Skills and Azure Functions Skills, see Azure Skills and Azure Functions Skills Boundary.

| Skill | Purpose | | --- | --- | | azure-functions-help | Discover and route to the best Azure Functions skill | | azure-functions-setup | Verify local prerequisites (Azure CLI, Core Tools, runtimes, Azure Skills) | | azure-functions-create | Create new Functions projects or add functions via Azure MCP templates | | azure-functions-agents | Build Azure Functions hosted AI agent apps, scheduled agents, connector-triggered agents, and chat/API agents | | azure-functions-deploy | Prepare, validate, and deploy via Azure Skills with Functions-specific guidance | | azure-functions-best-practices | Production-readiness review (config, security, reliability) | | azure-functions-diagnostics | Investigate deployment, runtime, trigger, binding, logging issues | | azure-functions-health-status | Collect current health, metrics, logs, Resource Health, Activity Log | | azure-functions-inventory | Collect app specification and configuration inventory | | azure-functions-doctor | Pre-deployment validation (used by the doctor CLI command) | | azure-functions-common | Shared language, trigger, binding, extension, routing references | | azure-functions-feedback | Turn session findings into previewed issues or pull requests |

The azure-functions-help skill provides the discovery and routing entry point.

Doctor — pre-deployment validation

Catch configuration mistakes, deprecated settings, and semantic code issues (missing error handling, blocking I/O, hardcoded secrets, durable-orchestrator non-determinism) before you deploy. The LLM semantic analysis is the value — doctor ships it as both a local CLI command and a GitHub Actions step.

Local — LLM analysis + visual HTML report

npx @azure/functions-skills doctor --dir . \
  --deep --accept-deep-risk \
  --agent github-copilot \
  --format html --output doctor-report.html

--accept-deep-risk acknowledges that the agent runs with elevated permissions (file write, shell execution) — only use on trusted workspaces. Skip the LLM with --no-deep for fast deterministic checks only.

Open doctor-report.html in a browser:

GitHub Actions — pre-deploy gate with deep analysis

Trigger on push: main (post-merge), not on pull requests — --deep refuses to run on pull-request workspaces because PR code is untrusted (it can prompt-inject the agent). See docs/doctor-guide.md#security-model.

on:
  push:
    branches: [main]

jobs:
  deep-doctor:
    runs-on: ubuntu-latest
    environment: trusted-deep-analysis  # GitHub Environment for approval + scoped secret
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '22'
      - name: Install GitHub Copilot CLI
        run: npm install -g @github/copilot
      - name: Run Azure Functions doctor
        env:
          GITHUB_TOKEN: ${{ secrets.COPILOT_TOKEN }}
        run: |
          npx @azure/functions-skills doctor \
            --deep --accept-deep-risk \
            --agent github-copilot \
            --format markdown --output doctor.md \
            --severity high
      - name: Publish summary
        if: always()
        run: cat doctor.md >> $GITHUB_STEP_SUMMARY

Exit code is 1 if any finding is at or above --severity (default high), gating downstream deploy steps. For PR validation, use the same command with --no-deep (Tier 1 only) on pull_request events.

Doctor walkthrough? See docs/doctor-guide.md for Tier 1 vs Tier 2 details, output formats, deep mode security, and bad-app fixtures.

Doctor also includes supply-chain security checks (lifecycle scripts, unpinned production dependencies, missing lockfile, tracked .env files, install-script deps, plus Tier 2 semantic checks for import-time side effects, fetch-then-execute, and credential exfiltration patterns) — informed by recent npm and PyPI compromises. See SECURITY.md for the threat model.

Contributing

We welcome contributions. The canonical source for skills, telemetry hooks, and MCP definitions lives under templates/ — edit there, then npm run build:plugin-payload to regenerate the published plugin payload.

Read CONTRIBUTING.md for the full guide.

Security

Report vulnerabilities to [email protected]. See SECURITY.md for the threat model and our defense layers.

License

MIT