{"id":30737704,"url":"https://github.com/inference-gateway/adl-cli","last_synced_at":"2026-05-23T10:03:30.432Z","repository":{"id":305548764,"uuid":"1023131482","full_name":"inference-gateway/adl-cli","owner":"inference-gateway","description":"A command-line tool to scaffold and manage enterprise-ready AI Agents powered by the A2A (Agent-to-Agent) protocol","archived":false,"fork":false,"pushed_at":"2026-01-27T18:23:19.000Z","size":825,"stargazers_count":7,"open_issues_count":0,"forks_count":2,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-01-28T04:00:40.994Z","etag":null,"topics":["a2a","agents","ai","containerized","docker","docker-compose","inference-gateway","metrics","observability","production-ready","scaffolding","tls-support","utils","vendor-agnostic"],"latest_commit_sha":null,"homepage":"https://github.com/inference-gateway","language":"Go","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/inference-gateway.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2025-07-20T15:39:09.000Z","updated_at":"2026-01-27T18:23:46.000Z","dependencies_parsed_at":"2025-09-30T19:11:04.882Z","dependency_job_id":"3863c778-deb4-4bcb-b94b-bdde641abf99","html_url":"https://github.com/inference-gateway/adl-cli","commit_stats":null,"previous_names":["inference-gateway/a2a-cli","inference-gateway/adl-cli"],"tags_count":126,"template":false,"template_full_name":null,"purl":"pkg:github/inference-gateway/adl-cli","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/inference-gateway%2Fadl-cli","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/inference-gateway%2Fadl-cli/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/inference-gateway%2Fadl-cli/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/inference-gateway%2Fadl-cli/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/inference-gateway","download_url":"https://codeload.github.com/inference-gateway/adl-cli/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/inference-gateway%2Fadl-cli/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31643431,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-10T07:40:12.752Z","status":"ssl_error","status_checked_at":"2026-04-10T07:40:11.664Z","response_time":98,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["a2a","agents","ai","containerized","docker","docker-compose","inference-gateway","metrics","observability","production-ready","scaffolding","tls-support","utils","vendor-agnostic"],"created_at":"2025-09-03T21:46:40.134Z","updated_at":"2026-05-23T10:03:30.420Z","avatar_url":"https://github.com/inference-gateway.png","language":"Go","funding_links":[],"categories":["🛠️ Tools \u0026 Libraries"],"sub_categories":["🏗️ Frameworks \u0026 SDKs"],"readme":"\u003cdiv align=\"center\"\u003e\n\n# ADL CLI\n\n_A command-line interface for generating enterprise-ready A2A (Agent-to-Agent) servers from Agent Definition Language (ADL) files._\n\n\u003e ⚠️ **Early Development Warning**: This project is in its early stages of development. Breaking changes are expected and acceptable until we reach a stable version. Use with caution in production environments.\n\n[![Go Version](https://img.shields.io/github/go-mod/go-version/inference-gateway/adl-cli?style=flat-square\u0026logo=go)](https://golang.org)\n[![License](https://img.shields.io/badge/license-MIT-green.svg?style=flat-square)](LICENSE)\n[![Build Status](https://img.shields.io/github/actions/workflow/status/inference-gateway/adl-cli/ci.yml?style=flat-square\u0026logo=github)](https://github.com/inference-gateway/adl-cli/actions)\n[![Go Report Card](https://goreportcard.com/badge/github.com/inference-gateway/adl-cli?style=flat-square)](https://goreportcard.com/report/github.com/inference-gateway/adl-cli)\n[![Release](https://img.shields.io/github/v/release/inference-gateway/adl-cli?style=flat-square\u0026logo=github)](https://github.com/inference-gateway/adl-cli/releases)\n\n\u003c/div\u003e\n\n## Table of Contents\n\n- [Overview](#overview)\n- [Installation](#installation)\n  - [Quick Install (Recommended)](#quick-install-recommended)\n  - [From Source](#from-source)\n  - [Using Go Install](#using-go-install)\n  - [Pre-built Binaries](#pre-built-binaries)\n- [Quick Start](#quick-start)\n- [Usage](#usage)\n  - [Commands](#commands)\n  - [Init Command](#init-command)\n  - [Generate Command](#generate-command)\n- [Agent Definition Language (ADL)](#agent-definition-language-adl)\n- [Generated Project Structure](#generated-project-structure)\n- [Sandbox Environments](#sandbox-environments)\n- [Enterprise Features](#enterprise-features)\n- [Artifacts Support](#artifacts-support)\n- [GitHub Issue Templates](#github-issue-templates)\n- [Examples](#examples)\n- [Template System \u0026 Architecture](#template-system--architecture)\n- [Customizing Generation with .adl-ignore](#customizing-generation-with-adl-ignore)\n- [Configurable Acronyms](#configurable-acronyms)\n- [Post-Generation Hooks](#post-generation-hooks)\n- [Development](#development)\n- [Roadmap](#roadmap)\n- [License](#license)\n- [Support](#support)\n\n## Overview\n\nThe ADL CLI helps you build enterprise-ready A2A agents quickly by generating complete project scaffolding from YAML-based Agent Definition Language (ADL) files. It eliminates boilerplate code and ensures consistent patterns across your agent implementations.\n\n### Key Features\n\n- 🚀 **Rapid Development** - Generate complete projects in seconds\n- 📋 **Schema-Driven** - Use YAML Agent Definition Language files (ADL) to define your agents\n- 🎯 **Enterprise Ready** - Single unified template with AI integration and enterprise features\n- 🔐 **Enterprise Features** - Authentication, SCM integration, and audit logging\n- 🛠️ **Smart Ignore** - Protect your implementations with .adl-ignore files\n- ✅ **Validation** - Built-in ADL schema validation\n- 🛠️ **Interactive Setup** - Guided project initialization with extensive CLI options\n- 🔗 **Structured Services** - Type-safe dependency injection with interfaces and factory functions\n- ⚙️ **Configuration Management** - Automatic environment variable mapping with proper naming conventions\n- 🔧 **CI/CD Generation** - Automatic GitHub Actions workflows with semantic-release CD pipelines\n- 🏗️ **Sandbox Environments** - Flox and DevContainer support for isolated development\n- 🎣 **Post-Generation Hooks** - Customize build, format, and test commands after generation\n- 🤖 **Multi-Provider AI** - OpenAI, Anthropic, DeepSeek, Ollama, Google, Mistral, and Groq support\n- 📁 **Artifacts Support** - Integrated filesystem and MinIO object storage for artifact management\n\n## Installation\n\n### Quick Install (Recommended)\n\nUse our install script to automatically download and install the latest binary:\n\n```bash\ncurl -fsSL https://raw.githubusercontent.com/inference-gateway/adl-cli/main/install.sh | bash\n```\n\nOr download and run the script manually:\n\n```bash\nwget https://raw.githubusercontent.com/inference-gateway/adl-cli/main/install.sh\nchmod +x install.sh\n./install.sh\n```\n\n**Install Options:**\n\n- Install specific version: `./install.sh --version v1.0.0`\n- Custom install directory: `INSTALL_DIR=~/bin ./install.sh`\n- Show help: `./install.sh --help`\n\n### From Source\n\n```bash\ngit clone https://github.com/inference-gateway/adl-cli.git\ncd adl-cli\ngo install .\n```\n\n### Using Go Install\n\n```bash\ngo install github.com/inference-gateway/adl-cli@latest\n```\n\n### Nix Flake\n\nRun the latest version directly without installing:\n\n```bash\nnix run github:inference-gateway/adl-cli\n```\n\nOr pin a specific version:\n\n```bash\nnix run github:inference-gateway/adl-cli/v0.27.13\n```\n\nBuild and add it to your profile:\n\n```bash\nnix profile install github:inference-gateway/adl-cli/v0.27.13\n```\n\nEnter a development shell with `go`, `go-task`, `golangci-lint`, `gopls`, and\n`goreleaser` available:\n\n```bash\nnix develop github:inference-gateway/adl-cli\n```\n\n### Flox\n\nPin `adl` to a specific version inside a [Flox](https://flox.dev) environment by\nadding it to your `.flox/env/manifest.toml`:\n\n```toml\n[install]\nadl.flake = \"github:inference-gateway/adl-cli/v0.31.0\"\n```\n\nThen activate the environment:\n\n```bash\nflox activate\n```\n\n### Pre-built Binaries\n\nDownload pre-built binaries from the [releases page](https://github.com/inference-gateway/adl-cli/releases).\n\n## Quick Start\n\n### 1. Initialize a New Project\n\n```bash\n# Interactive project setup - creates ADL manifest\nadl init my-weather-agent\n\n# Generate project code from the manifest\nadl generate --file agent.yaml --output ./test-my-agent\n```\n\n### 2. Implement Your Business Logic\n\nThe generated project includes TODO placeholders for your implementations:\n\n```go\n// TODO: Implement weather API logic\nfunc GetWeatherTool(ctx context.Context, args map[string]any) (string, error) {\n    city := args[\"city\"].(string)\n    // TODO: Replace with actual weather API call\n    return fmt.Sprintf(`{\"city\": \"%s\", \"temp\": \"22°C\"}`, city), nil\n}\n```\n\n### 3. Build and Run\n\n```bash\ncd test-weather-agent\ntask build\ntask run\n```\n\n## Usage\n\n### Commands\n\n| Command               | Description                                                        |\n| --------------------- | ------------------------------------------------------------------ |\n| `adl init [name]`     | Create ADL manifest file interactively with options                |\n| `adl generate`        | Generate project code from ADL file with CI/CD and sandbox support |\n| `adl validate [file]` | Validate an ADL file against the complete schema                   |\n\n### Init Command\n\nThe `adl init` command provides a interactive wizard for creating ADL manifest files:\n\n```bash\n# Interactive ADL manifest creation\nadl init my-weather-agent\n\n# Use defaults for all prompts\nadl init my-agent --defaults\n\n# Non-interactive with specific configuration\nadl init my-agent \\\n  --name \"Weather Agent\" \\\n  --description \"Provides weather information\" \\\n  --provider deepseek \\\n  --model deepseek-v4-flash \\\n  --language go \\\n  --flox\n```\n\n#### Init Command Options\n\nThe init command supports extensive configuration options:\n\n**Project Settings:**\n\n- `--defaults` - Use default values for all prompts\n- `--path` - Project directory path\n- `--name` - Agent name\n- `--description` - Agent description\n- `--version` - Agent version\n\n**Agent Configuration:**\n\n- `--type` - Agent type (`ai-powered`/`minimal`)\n- `--provider` - AI provider (`openai`/`anthropic`/`deepseek`/`ollama`/`google`/`mistral`/`groq`)\n- `--model` - AI model name\n- `--system-prompt` - System prompt for the agent\n- `--max-tokens` - Maximum tokens (integer)\n- `--temperature` - Temperature (0.0-2.0)\n\n**Capabilities:**\n\n- `--streaming` - Enable streaming responses\n- `--notifications` - Enable push notifications\n- `--history` - Enable state transition history\n\n**Server Configuration:**\n\n- `--port` - Server port (integer)\n- `--debug` - Enable debug mode\n\n**Language-Specific Options:**\n\n- `--language` - Programming language (`go`/`rust`, TypeScript support planned)\n\n**Go Options:**\n\n- `--go-module` - Go module path (e.g., `github.com/user/project`)\n- `--go-version` - Go version (e.g., `1.26.2`)\n\n**Rust Options:**\n\n- `--rust-package-name` - Rust package name\n- `--rust-version` - Rust version (e.g., `1.94`)\n- `--rust-edition` - Rust edition (e.g., `2024`)\n\n**TypeScript Options:**\n\n- `--typescript-name` - TypeScript package name\n\n**Environment Options:**\n\n- `--flox` - Enable Flox environment\n- `--devcontainer` - Enable DevContainer environment\n\n**Pipeline / AI Options (declarative, written into the manifest as `false` by default):**\n\n- `--ai` - Shortcut for the init wizard: writes\n  `spec.development.ai.claudecode.enabled: true` into the generated `agent.yaml`.\n  Every other per-agent toggle (`codex`, `gemini`, `opencode`, `infer`) stays\n  off; edit `agent.yaml` after init to enable additional agents\n  (see [Per-agent AI assistants](#per-agent-ai-assistants)).\n- `--ci` - Sets `spec.scm.ci: true` (generate CI workflow on `adl generate`)\n- `--cd` - Sets `spec.scm.cd: true` (generate CD pipeline + semantic-release on `adl generate`)\n\n### Generate Command\n\n```bash\n# Generate project from ADL file\nadl generate --file agent.yaml --output ./test-my-agent\n\n# Overwrite existing files (respects .adl-ignore)\nadl generate --file agent.yaml --output ./test-my-agent --overwrite\n\n# Generate with CI workflow configuration\nadl generate --file agent.yaml --output ./test-my-agent --ci\n\n# Generate with CloudRun deployment configuration\nadl generate --file agent.yaml --output ./test-my-agent --deployment cloudrun\n\n# Generate with CloudRun deployment and CD pipeline\nadl generate --file agent.yaml --output ./test-my-agent --deployment cloudrun --cd\n```\n\n#### Generate Flags\n\n| Flag               | Description                                                                                |\n| ------------------ | ------------------------------------------------------------------------------------------ |\n| `--file`, `-f`     | ADL file to generate from (default: \"agent.yaml\")                                          |\n| `--output`, `-o`   | Output directory for generated code (default: \".\")                                         |\n| `--template`, `-t` | Template to use (default: \"minimal\")                                                       |\n| `--overwrite`      | Overwrite existing files (respects .adl-ignore)                                            |\n| `--ci`             | Generate CI workflow configuration (GitHub Actions). Overrides `spec.scm.ci`.              |\n| `--cd`             | Generate CD pipeline configuration with semantic-release. Overrides `spec.scm.cd`.         |\n| `--deployment`     | Generate deployment configuration (`kubernetes`, `cloudrun`)                               |\n\n\u003e **Declarative equivalents:** `--ci` and `--cd` are mirrored by `spec.scm.ci`\n\u003e and `spec.scm.cd`. The CLI flag is OR'd on top of the manifest value (passing\n\u003e the flag wins; omitting it falls back to the manifest). AI assistants are\n\u003e entirely manifest-driven via the per-agent toggles in `spec.development.ai`\n\u003e - see the matrix below.\n\u003e `adl init` writes all toggles as `false` by default - they're opt-in. Generated files\n\u003e (`CLAUDE.md`, `GEMINI.md`, `AGENTS.md`, `.github/workflows/ci.yml`,\n\u003e `.github/workflows/cd.yml`, `.github/workflows/claude.yml`,\n\u003e `.github/workflows/codex.yml`, `.github/workflows/gemini.yml`,\n\u003e `.releaserc.yaml`) are tagged `linguist-generated=true` in `.gitattributes`\n\u003e so they collapse in pull request diffs.\n\n**CI Generation Features:**\n\n- **Automatic Provider Detection**: Detects GitHub from ADL `spec.scm.provider` (GitLab support planned)\n- **Language-Specific Workflows**: Tailored CI configurations for Go, Rust, and TypeScript\n- **Version Integration**: Uses language versions from ADL configuration\n- **Task Integration**: Leverages generated Taskfile for consistent build processes\n- **Caching**: Includes service caching for faster builds\n\n**CD Generation Features:**\n\n- **Semantic Release Integration**: Automatic versioning based on conventional commits\n- **Multi-Language Support**: Builds and tests for Go, Rust, and TypeScript projects\n- **Container Publishing**: Builds and pushes Docker images to GitHub Container Registry\n- **Manual Dispatch**: CD workflow triggered manually via GitHub Actions\n- **Changelog Generation**: Automatic CHANGELOG.md generation with release notes\n- **GitHub Releases**: Creates GitHub releases with appropriate tagging\n- **Deployment Integration**: Supports automatic deployment to Kubernetes and Cloud Run after successful releases\n\n**AI Integration Features:**\n\nThe ADL CLI honours the per-agent toggles in `spec.development.ai` (introduced\nin ADL schema v0.8.0). Each entry is independent and defaults to `false`:\n\n```yaml\nspec:\n  development:\n    ai:\n      claudecode:\n        enabled: true   # generates CLAUDE.md + .github/workflows/claude.yml\n      codex:\n        enabled: false  # would generate AGENTS.md + .github/workflows/codex.yml\n      gemini:\n        enabled: false  # would generate GEMINI.md + .github/workflows/gemini.yml\n      opencode:\n        enabled: false  # would generate AGENTS.md (no upstream action yet)\n      infer:\n        enabled: false  # would generate AGENTS.md (no upstream action yet)\n```\n\n#### Per-agent AI assistants\n\n| Agent toggle  | Docs file the agent reads | GitHub Actions workflow generated? |\n|---------------|---------------------------|------------------------------------|\n| `claudecode`  | `CLAUDE.md`               | yes (`.github/workflows/claude.yml`, uses `anthropics/claude-code-action`) |\n| `codex`       | `AGENTS.md` (shared)      | yes (`.github/workflows/codex.yml`, uses `openai/codex-action`) |\n| `gemini`      | `GEMINI.md`               | yes (`.github/workflows/gemini.yml`, uses `google-github-actions/run-gemini-cli`) |\n| `opencode`    | `AGENTS.md` (shared)      | no upstream action yet - docs only |\n| `infer`       | `AGENTS.md` (shared)      | no workflow scaffolded yet - docs only |\n\n- `AGENTS.md` is generated **once** and is shared by every enabled agent that\n  reads from it (`codex`, `opencode`, `infer`); the file's contents are\n  agent-agnostic.\n- `CLAUDE.md` and `GEMINI.md` are agent-specific and only appear when the\n  matching toggle is on.\n- If no toggles are enabled, no AI docs or workflows are emitted.\n- Pre-v0.8.0 manifests using `spec.development.ai.enabled: true` are no longer\n  accepted - `adl validate` and `adl generate` will fail with a migration hint\n  pointing at the per-agent toggles. Move `enabled: true` to the specific agent\n  you want (e.g. `claudecode.enabled: true`).\n- When `claudecode` is enabled, sandbox environments (Flox, DevContainer)\n  also gain the `claude-code` CLI / extension automatically.\n\n**Deployment Generation Features:**\nThe `--deployment` flag generates platform-specific deployment configurations:\n\n- **CloudRun Deployment**: Creates a `deploy` task in the root `Taskfile.yml` for gcloud deployment\n  - Supports both Google Container Registry (GCR) and GitHub Container Registry (GHCR)\n  - Configurable resources (CPU, memory), scaling (min/max instances), and service options\n  - Uses direct gcloud commands for truly serverless deployment (no Kubernetes required)\n  - Automatic container building with Docker or Cloud Build integration\n- **Kubernetes Deployment**: Creates `k8s/deployment.yaml` with standard Kubernetes manifests\n  - Enterprise-ready configurations with resource limits and health checks\n  - ConfigMap and Secret integration for environment variables\n  - Service and Ingress configurations for load balancing\n\n## Agent Definition Language (ADL)\n\nADL files use YAML to define your agent's configuration, capabilities, and tools.\n\nThe canonical schema lives in the [inference-gateway/adl](https://github.com/inference-gateway/adl) repository - that repo is the single source of truth for the ADL specification. This CLI vendors a pinned copy at `internal/schema/schema.json` (refresh with `task fetch-schema`).\n\n### Example ADL File\n\n```yaml\napiVersion: adl.inference-gateway.com/v1\nkind: Agent\nmetadata:\n  name: weather-agent\n  description: \"Provides weather information for cities worldwide\"\n  version: \"1.0.0\"\nspec:\n  capabilities:\n    streaming: true\n    pushNotifications: false\n    stateTransitionHistory: false\n  agent:\n    provider: \"\" # Choose: openai, anthropic, deepseek, ollama, google, mistral, groq\n    model: \"\" # Specify default model name for chosen provider\n    systemPrompt: \"You are a helpful weather assistant.\"\n    maxTokens: 4096\n    temperature: 0.7\n  tools:\n    - name: get_weather\n      description: \"Get current weather for a city\"\n      schema:\n        type: object\n        properties:\n          city:\n            type: string\n            description: \"City name\"\n          country:\n            type: string\n            description: \"Country code\"\n        required:\n          - city\n  server:\n    port: 8080\n    debug: false\n  language:\n    go:\n      module: \"github.com/example/weather-agent\"\n      version: \"1.26.2\"\n  acronyms: # Optional: Custom acronyms for better code generation\n    - api\n    - json\n    - xml\n```\n\n### ADL Schema\n\nThe complete ADL schema includes:\n\n- **metadata**: Agent name, description, and version\n- **capabilities**: Streaming, notifications, state history\n- **config**: Structured configuration sections with environment variable mapping\n- **services**: Service services with interfaces, factories, and type definitions\n- **agent**: AI provider configuration (OpenAI, Anthropic, DeepSeek, Ollama, Google, Mistral, Groq)\n- **tools**: Function-call definitions with JSON schemas, validation, and service injection support\n- **skills**: Markdown playbooks (id + optional `bare`, version, source) pulled from the skills registry, fetched as a full directory from a GitHub repo (shorthand or URL), or scaffolded locally; advertised on the agent card and prepended to the system prompt at runtime\n- **server**: HTTP server configuration with authentication support\n- **language**: Programming language-specific settings (Go, Rust, TypeScript) and configurable acronyms\n- **scm**: Source control management configuration (GitHub, GitLab)\n- **sandbox**: Development environment configuration (Flox, DevContainer)\n- **deployment**: Platform-specific deployment configuration (Kubernetes, Cloud Run)\n\n### Complete ADL Example\n\n```yaml\napiVersion: adl.inference-gateway.com/v1\nkind: Agent\nmetadata:\n  name: advanced-agent\n  description: \"Enterprise agent with full feature set\"\n  version: \"1.0.0\"\nspec:\n  capabilities:\n    streaming: true\n    pushNotifications: true\n    stateTransitionHistory: true\n  agent:\n    provider: deepseek\n    model: deepseek-v4-flash\n    systemPrompt: |\n      You are a helpful assistant with enterprise capabilities.\n      Always prioritize security and compliance.\n    maxTokens: 8192\n    temperature: 0.3\n  config:\n    database:\n      connectionString: \"postgresql://user:pass@localhost:5432/db\"\n      maxConnections: \"10\"\n      timeout: \"30s\"\n    notifications:\n      slackWebhook: \"https://hooks.slack.com/services/...\"\n      emailApiKey: \"your-email-api-key\"\n      retryAttempts: \"3\"\n  services:\n    database:\n      type: service\n      interface: DatabaseService\n      factory: NewDatabaseService\n      description: PostgreSQL database service for persistent storage\n    notifications:\n      type: service\n      interface: NotificationService\n      factory: NewNotificationService\n      description: Multi-channel notification service\n  tools:\n    - name: query_database\n      description: \"Execute database queries with validation\"\n      inject:\n        - logger\n        - database\n      schema:\n        type: object\n        properties:\n          query:\n            type: string\n            description: \"SQL query to execute\"\n          table:\n            type: string\n            description: \"Target table name\"\n          limit:\n            type: integer\n            description: \"Result limit\"\n            maximum: 1000\n        required:\n          - query\n          - table\n    - name: send_notification\n      description: \"Send multi-channel notifications\"\n      inject:\n        - logger\n        - notifications\n      schema:\n        type: object\n        properties:\n          recipient:\n            type: string\n            description: \"Recipient identifier\"\n          message:\n            type: string\n            description: \"Message content\"\n          priority:\n            type: string\n            enum:\n              - low\n              - medium\n              - high\n              - critical\n          channel:\n            type: string\n            enum:\n              - email\n              - slack\n              - teams\n              - webhook\n        required:\n          - recipient\n          - message\n          - priority\n          - channel\n  server:\n    port: 8443\n    debug: false\n    auth:\n      enabled: true\n  language:\n    go:\n      module: \"github.com/company/advanced-agent\"\n      version: \"1.26.2\"\n  scm:\n    provider: github\n    url: \"https://github.com/company/advanced-agent\"\n  deployment:\n    type: cloudrun\n    cloudrun:\n      image:\n        registry: gcr.io\n        repository: advanced-agent\n        tag: latest\n        useCloudBuild: true\n      resources:\n        cpu: \"2\"\n        memory: 1Gi\n      scaling:\n        minInstances: 1\n        maxInstances: 100\n        concurrency: 1000\n      service:\n        timeout: 3600\n        allowUnauthenticated: false\n        serviceAccount: advanced-agent@PROJECT_ID.iam.gserviceaccount.com\n        executionEnvironment: gen2\n      environment:\n        LOG_LEVEL: info\n        ENVIRONMENT: production\n  development:\n    sandbox:\n      flox:\n        enabled: true\n```\n\n## Skills vs. Tools\n\nThe ADL spec distinguishes two complementary concepts:\n\n- **Tools** (`spec.tools`) are function-call entrypoints with explicit JSON schemas. They are generated as code in the target language and registered with the agent's toolbox. The model invokes them by name with structured arguments.\n- **Skills** (`spec.skills`) are markdown playbooks (with YAML frontmatter) that describe _when and how_ to use the tools. Each is written to its own directory at `skills/\u003cid\u003e/SKILL.md` in the generated project, advertised on the agent card so orchestrators can discover them, and prepended to the system prompt at runtime. The directory layout matches Anthropic's [agent skills convention](https://github.com/anthropics/skills) - bare skills can ship arbitrary scripts, templates, or reference material alongside `SKILL.md`.\n\nA skill entry is small:\n\n```yaml\nspec:\n  skills:\n    - id: data-analysis          # pulled from registry.inference-gateway.com/skills/\n      version: 0.1.0             # optional pin\n    - id: report-writing         # pulled at the default version\n    - id: company-policy         # scaffolded locally (not fetched)\n      bare: true\n      name: company-policy\n      description: \"Internal compliance rules to follow\"\n      license: Proprietary       # optional SPDX id or \"Proprietary\"\n      tags: [policy, compliance]\n    - id: pdf                    # pull a full skill directory from GitHub\n      source: anthropics/skills/pdf\n    - id: skill-creator\n      source: skill-creator@v1.0 # pin to a tag/branch/sha\n```\n\n### Resolution rules\n\n- **`bare: true`** → the CLI scaffolds `skills/\u003cid\u003e/SKILL.md` with frontmatter from the manifest and a TODO body that you author by hand. The whole `skills/\u003cid\u003e/` directory is listed in `.adl-ignore`, so any bundled scripts, templates, or resources you drop alongside `SKILL.md` are preserved on regeneration.\n- **`source:` set** → the source must resolve to a public GitHub directory (a `/tree/\u003cref\u003e/\u003cpath\u003e` URL, or one of the shorthand forms below). The CLI pulls the _entire_ directory - `SKILL.md`, reference docs, bundled scripts, anything else - and writes it to `skills/\u003cid\u003e/`. Non-`github.com` URLs are rejected so the same code path always produces a complete skill bundle, not a stray markdown file.\n- **Otherwise** → fetch `https://registry.inference-gateway.com/skills/\u003cid\u003e[/\u003cversion\u003e].md` (becomes `skills/\u003cid\u003e/SKILL.md`). Override the registry with `ADL_SKILLS_REGISTRY`. Registry-by-id currently ships `SKILL.md` only; if you need bundled assets, use `source:` to point at a GitHub directory.\n\n### Licensing\n\n`license` is optional on every skill entry. When set, it must be one of the\nSPDX identifiers enumerated in the schema (`MIT`, `Apache-2.0`, `BSD-2-Clause`,\n`BSD-3-Clause`, `GPL-2.0`, `GPL-3.0`, `LGPL-2.1`, `LGPL-3.0`, `MPL-2.0`, `ISC`,\n`CC0-1.0`, `CC-BY-4.0`, `CC-BY-SA-4.0`, `Unlicense`) or the literal string\n`Proprietary` for closed-source skills. The resolver mirrors the value into the\ngenerated `SKILL.md` frontmatter so the licence travels with the playbook -\nshipping a separate `LICENSE` file alongside `SKILL.md` is optional. When the\nADL entry and the fetched frontmatter both set `license`, the value in the ADL\nmanifest wins.\n\nUse `adl generate --offline` to skip network access - every non-bare skill must already be cached at `~/.adl/skills-cache/\u003cid\u003e@\u003cref\u003e/` (where `\u003cref\u003e` is the pinned tag/branch, or `latest` for an unpinned registry fetch).\n\n### `source:` shorthand grammar\n\nEvery form below resolves to a GitHub `tree/\u003cref\u003e/\u003cpath\u003e` URL. An optional `@\u003ctag\u003e` suffix pins a branch, tag, or commit SHA; omit it to use the default `main` branch.\n\n| Shorthand                        | Expands to                                                                   |\n| -------------------------------- | ---------------------------------------------------------------------------- |\n| `\u003cskill\u003e`                        | `https://github.com/inference-gateway/skills/tree/main/skills/\u003cskill\u003e`       |\n| `\u003cskill\u003e@\u003ctag\u003e`                  | `https://github.com/inference-gateway/skills/tree/\u003ctag\u003e/skills/\u003cskill\u003e`      |\n| `\u003cowner\u003e/\u003crepo\u003e/\u003cskill\u003e`         | `https://github.com/\u003cowner\u003e/\u003crepo\u003e/tree/main/skills/\u003cskill\u003e`                 |\n| `\u003cowner\u003e/\u003crepo\u003e/\u003cskill\u003e@\u003ctag\u003e`   | `https://github.com/\u003cowner\u003e/\u003crepo\u003e/tree/\u003ctag\u003e/skills/\u003cskill\u003e`                |\n| Full `https://github.com/...` URL | passed through unchanged                                                    |\n\nConcrete examples:\n\n```yaml\n# Default inference-gateway/skills, latest main:\n- id: skill-creator\n  source: skill-creator\n\n# Default inference-gateway/skills, pinned to a tag:\n- id: skill-creator\n  source: skill-creator@v1.0\n\n# Different repo (Anthropic's official skill library):\n- id: pdf\n  source: anthropics/skills/pdf\n\n# Different repo, pinned to a commit SHA:\n- id: pdf\n  source: anthropics/skills/pdf@abc1234\n\n# Full URL for anything that doesn't fit the shorthand:\n- id: custom\n  source: https://github.com/my-org/my-repo/tree/release/path/to/skill\n```\n\nThe 3-segment form assumes a `skills/\u003cid\u003e/` subdirectory inside the repo (the convention used by both `inference-gateway/skills` and `anthropics/skills`). If your repo lays skills out differently, pass the full URL.\n\n### Runtime: AVAILABLE SKILLS manifest + on-demand Read\n\nThe generated agent advertises skills to the LLM via a frontmatter-only manifest, **not** by inlining SKILL.md bodies. At startup it walks first-level subdirectories under `skills/` (overridable with `A2A_SKILLS_DIR`), parses each `\u003cid\u003e/SKILL.md`'s YAML frontmatter, and appends an `AVAILABLE SKILLS:` block to the system prompt:\n\n```text\nAVAILABLE SKILLS:\nSkills are reusable instructions for specific tasks. When a task matches a\nskill's description, read the SKILL.md file at the listed path using the Read\ntool, then follow its instructions.\n\n- incident-response: Use this when the user reports a production incident...\n  Path: skills/incident-response/SKILL.md\n- pdf: Fill in PDF forms and extract structured data from PDFs.\n  Path: skills/pdf/SKILL.md\n```\n\nThe model loads each SKILL.md body on demand via the `Read` built-in tool, and executes any bundled scripts via `Bash` / `Write` / `Edit`. **A skills-using agent must therefore list `- id: read` in `spec.tools` and set `spec.config.tools.read.enabled: true`** - the validator enforces this; see [Reserved built-in tools](#reserved-built-in-tools).\n\n### Reserved built-in tools\n\n`spec.tools` accepts five reserved IDs that map to framework-supplied implementations:\n\n| Reserved ID | Generated as            | Purpose                                                                              |\n| ----------- | ----------------------- | ------------------------------------------------------------------------------------ |\n| `read`      | `tools/read.go` etc.    | Read a file (`file_path`, optional `offset`/`limit`).                                |\n| `bash`      | `tools/bash.go` etc.    | Execute a shell command (subject to whitelist + timeout).                            |\n| `write`     | `tools/write.go` etc.   | Write content to a file (creates parent dirs).                                       |\n| `edit`      | `tools/edit.go` etc.    | Replace a unique string in a file (`old_string` → `new_string`).                     |\n| `fetch`     | `tools/fetch.go` etc.   | Fetch an http(s) URL (whitelist, max-bytes cap, optional save-to-disk inside `/tmp`).|\n\nOpt in by listing the id alone - the generator owns `name`, `description`, and the JSON schema:\n\n```yaml\nspec:\n  tools:\n    - id: read\n    - id: bash\n    - id: query_database       # user tool: full entry still required\n      name: query_database\n      description: \"...\"\n      schema: { type: object, ... }\n```\n\n**All five built-ins default to `enabled: false`.** Activate them via the reserved namespace `spec.config.tools.\u003cid\u003e`:\n\n```yaml\nspec:\n  config:\n    tools:\n      read:\n        enabled: true\n        max_lines: 2000          # offset/limit default window\n        allowed_roots: []        # empty = project-wide\n      bash:\n        enabled: true\n        whitelist: [ls, cat, grep, jq]\n        timeout_seconds: 30\n      write:\n        enabled: false           # listed but explicitly disabled\n      edit:\n        enabled: true\n      fetch:\n        enabled: true\n        allowed_domains:         # whitelist of hosts (empty = unrestricted, discouraged)\n          - example.com\n          - .api.dev             # entries starting with \".\" match any subdomain\n        max_bytes: 10485760      # 10 MiB cap on response body (default)\n        timeout_seconds: 30      # total request timeout (default)\n        download_dir: /tmp       # root for save_path writes (default /tmp)\n        allow_downloads: false   # set true to allow writing response bodies to disk\n```\n\nValues are baked into the generated constructor as compile-time literals - there's no `ToolsConfig` struct in `config/config.go` because reserved-namespace sections are intentionally skipped. The validator decodes each `spec.config.tools.\u003cid\u003e` block into the built-in's typed shape and rejects unknown keys (typos like `tymeout_seconds` fail with `spec.config.tools.bash.tymeout_seconds`).\n\nRuntime overrides for Bash (read inside `tools/bash.go`):\n\n- `A2A_BASH_DISABLED=1` is a kill switch - overrides `enabled: true` back to false.\n- `A2A_BASH_WHITELIST=ls,cat,grep` overrides the compile-time whitelist.\n\nRuntime overrides for Fetch (resolution precedence: env \u003e compile-time literal \u003e default-disabled):\n\n- Go: `TOOLS_FETCH_ENABLED`, `TOOLS_FETCH_ALLOWED_DOMAINS`, `TOOLS_FETCH_MAX_BYTES`, `TOOLS_FETCH_TIMEOUT_SECONDS`, `TOOLS_FETCH_DOWNLOAD_DIR`, `TOOLS_FETCH_ALLOW_DOWNLOADS` (envconfig-style; comma-separated for lists).\n- Rust: `A2A_FETCH_DISABLED=1` (kill switch), `A2A_FETCH_ALLOWED_DOMAINS`, `A2A_FETCH_MAX_BYTES`, `A2A_FETCH_TIMEOUT_SECONDS`, `A2A_FETCH_DOWNLOAD_DIR`, `A2A_FETCH_ALLOW_DOWNLOADS`.\n\nThe Fetch tool supports `GET` and `HEAD` only. Optional `save_path` writes the response body to a path resolved under `download_dir` - absolute paths and parent-directory traversal (`..`) are rejected, and the request fails unless `allow_downloads: true`. Bodies (and on-disk files) are capped at `max_bytes`; oversized responses are truncated and the result payload sets `\"truncated\": true`. The Go template uses only the standard library (`net/http`); the Rust template adds `reqwest` (rustls-tls + json features) to `Cargo.toml` automatically when `- id: fetch` is present in `spec.tools`.\n\nResolution precedence at runtime: **env \u003e compile-time literal \u003e built-in default (disabled)**.\n\n## Service Injection \u0026 Configuration Management\n\nThe ADL CLI provides a sophisticated service injection system with structured configuration management. This system improves testability, separation of concerns, and provides type-safe configuration with environment variable mapping.\n\n### Structured Service System\n\nDefine services with explicit types, interfaces, and factory functions. The system supports both built-in services (like logger) and custom service services:\n\n```yaml\nspec:\n  config:\n    googleCalendar:\n      scopes: \"https://www.googleapis.com/auth/calendar\"\n      credentialsPath: \"/secrets/credentials.json\"\n    cache:\n      ttl: \"3600\"\n      maxEntries: \"1000\"\n  services:\n    googleCalendar:\n      type: service\n      interface: CalendarService\n      factory: NewCalendarService\n      description: Google Calendar API service for managing calendar events\n    cache:\n      type: service\n      interface: CacheRepository\n      factory: NewCacheRepository\n      description: High-performance caching layer for API responses\n  tools:\n    - name: create_event\n      description: \"Create a new calendar event\"\n      inject:\n        - logger # Built-in, always available\n        - googleCalendar # Custom service\n        - cache # Custom service\n      schema:\n        type: object\n        properties:\n          title:\n            type: string\n            description: \"Event title\"\n          start:\n            type: string\n            description: \"Start time (ISO 8601)\"\n        required: [title, start]\n```\n\n### Configuration Management\n\nThe configuration system generates type-safe structs with automatic environment variable mapping:\n\n**Generated Configuration (`config/config.go`):**\n\n```go\ntype Config struct {\n    // Core application settings\n    Environment string `env:\"ENVIRONMENT\"`\n\n    // A2A configuration\n    A2A serverConfig.Config `env:\",prefix=A2A_\"`\n\n    // Custom configuration sections\n    Cache          CacheConfig          `env:\",prefix=CACHE_\"`\n    GoogleCalendar GoogleCalendarConfig `env:\",prefix=GOOGLE_CALENDAR_\"`\n}\n\ntype GoogleCalendarConfig struct {\n    CredentialsPath string `env:\"CREDENTIALS_PATH\"`\n    Scopes          string `env:\"SCOPES\"`\n}\n\ntype CacheConfig struct {\n    MaxEntries string `env:\"MAX_ENTRIES\"`\n    Ttl        string `env:\"TTL\"`\n}\n```\n\n**Environment Variables:**\n\n- `GOOGLE_CALENDAR_CREDENTIALS_PATH=\"/secrets/google-creds.json\"`\n- `GOOGLE_CALENDAR_SCOPES=\"https://www.googleapis.com/auth/calendar\"`\n- `CACHE_MAX_ENTRIES=\"1000\"`\n- `CACHE_TTL=\"3600\"`\n\n### Config Subsection Injection\n\nIn addition to injecting entire configuration objects, you can inject specific config subsections directly into skills using dotted notation. This provides type-safe access to focused configuration scopes.\n\n**Example ADL Configuration:**\n\n```yaml\nspec:\n  config:\n    database:\n      connectionString: \"postgresql://localhost:5432/db\"\n      maxConnections: \"10\"\n      timeout: \"30s\"\n    email:\n      apiKey: \"\"\n      fromAddress: \"noreply@example.com\"\n      provider: \"sendgrid\"\n  services:\n    database:\n      type: service\n      interface: DatabaseService\n      factory: NewDatabaseService\n      description: PostgreSQL database service\n  tools:\n    - name: export_report\n      description: \"Export data and email report\"\n      inject:\n        - logger\n        - database\n        - config.email  # Inject only the email config subsection\n      schema:\n        type: object\n        properties:\n          recipient:\n            type: string\n        required: [recipient]\n```\n\n**Generated Skill Code:**\n\n```go\ntype ExportReportSkill struct {\n    logger   *zap.Logger\n    database database.DatabaseService\n    email    *config.EmailConfig  // Type-safe access to email config only\n}\n\nfunc NewExportReportSkill(\n    logger *zap.Logger,\n    database database.DatabaseService,\n    email *config.EmailConfig,\n) server.Tool {\n    skill := \u0026ExportReportSkill{\n        logger:   logger,\n        database: database,\n        email:    email,\n    }\n    // ...\n}\n\nfunc (s *ExportReportSkill) ExportReportHandler(ctx context.Context, args map[string]any) (string, error) {\n    // Direct access to email config subsection\n    apiKey := s.email.APIKey\n    fromAddress := s.email.FromAddress\n    provider := s.email.Provider\n\n    // ... implementation\n}\n```\n\n**Main Registration:**\n\n```go\n// In main.go - config subsection is passed directly\nexportReportSkill := skills.NewExportReportSkill(l, databaseSvc, \u0026cfg.Email)\ntoolBox.AddTool(exportReportSkill)\n```\n\n**Benefits of Config Subsection Injection:**\n\n- **Scoped Access**: Skills only receive the configuration they need, following principle of least privilege\n- **Type Safety**: Compile-time validation ensures config fields exist\n- **Clear Dependencies**: Explicit declaration of which config sections each skill requires\n- **Easier Testing**: Mock specific config subsections without full config object\n- **Better Separation**: Skills don't have access to unrelated configuration\n- **Auto-Validation**: ADL CLI validates that injected config sections exist in `spec.config`\n\n**Injection Patterns:**\n\n```yaml\ninject:\n  - logger                 # Built-in logger service\n  - config                 # Entire config object (*config.Config)\n  - config.database        # Database config subsection (*config.DatabaseConfig)\n  - config.email           # Email config subsection (*config.EmailConfig)\n  - myService              # Custom service from spec.services\n```\n\n### Service Architecture\n\nThe service injection system generates:\n\n1. **Built-in Logger**: Automatically available as `*zap.Logger` without declaration\n2. **Type-Safe Configuration**: Structured config with environment variable mapping\n3. **Service Interfaces**: Custom service packages with interface definitions\n4. **Factory Functions**: Constructor functions that receive logger and configuration\n5. **Automatic Registration**: Services are automatically wired into skills\n6. **File Protection**: Generated service files are automatically added to `.adl-ignore`\n\n### Generated Structure\n\n```text\nmy-agent/\n├── config/\n│   └── config.go                    # Type-safe configuration with env mapping\n├── internal/\n│   ├── logger/\n│   │   └── logger.go               # Built-in logger factory\n│   ├── googleCalendar/\n│   │   └── googleCalendar.go       # Calendar service with interface\n│   └── cache/\n│       └── cache.go                # Cache service with interface\n├── tools/\n│   ├── create_event.go             # Function-call tools with injected services\n│   └── list_events.go\n├── skills/\n│   ├── calendar-workflow/          # Markdown playbooks loaded into the system prompt\n│   │   └── SKILL.md\n│   └── meeting-summary/\n│       └── SKILL.md\n└── .adl-ignore                     # Protects custom implementations\n```\n\n### Generated Service Code\n\nEach service generates a package with interface and factory:\n\n**Example `internal/googleCalendar/googleCalendar.go`:**\n\n```go\ntype CalendarService interface {\n    // TODO: Define your CalendarService interface methods\n    CreateEvent(ctx context.Context, event *Event) error\n    ListEvents(ctx context.Context, query *Query) ([]*Event, error)\n}\n\ntype calendarService struct {\n    logger *zap.Logger\n    config *config.Config\n}\n\nfunc NewCalendarService(logger *zap.Logger, cfg *config.Config) (CalendarService, error) {\n    // TODO: Implement CalendarService initialization\n    return \u0026calendarService{\n        logger: logger,\n        config: cfg,\n    }, nil\n}\n```\n\n### Skill Integration\n\nSkills automatically receive injected services as constructor parameters:\n\n**Example `skills/create_event.go`:**\n\n```go\ntype CreateEventSkill struct {\n    logger    *zap.Logger\n    calendar  googleCalendar.CalendarService\n    cache     cache.CacheRepository\n}\n\nfunc NewCreateEventSkill(logger *zap.Logger, calendar googleCalendar.CalendarService, cache cache.CacheRepository) *CreateEventSkill {\n    return \u0026CreateEventSkill{\n        logger:   logger,\n        calendar: calendar,\n        cache:    cache,\n    }\n}\n```\n\n### Benefits\n\n- **Type Safety**: Structured configuration with compile-time validation\n- **Environment Variables**: Automatic mapping with proper naming conventions\n- **Interface-Based Design**: Testable services with clear contracts\n- **Separation of Concerns**: Configuration separate from service definitions\n- **Language Agnostic**: Works across Go, Rust, and planned TypeScript support\n- **Hot Reload**: Configuration changes via environment variables\n- **Security**: No secrets in code, environment-based configuration\n- **Scalability**: Easy to add new services and configuration sections\n\n### Best Practices\n\n1. **Configuration**: Use environment variables for secrets and environment-specific values\n2. **Interfaces**: Define clear interfaces for testability and modularity\n3. **Factory Functions**: Initialize services with proper error handling\n4. **Logging**: Use the injected logger for consistent log formatting\n5. **Testing**: Create mock implementations of service interfaces\n6. **Documentation**: Document interface methods and configuration options\n\n## Generated Project Structure\n\nThe ADL CLI generates project scaffolding tailored to your chosen language:\n\n### Go Project Structure\n\n```text\nmy-go-agent/\n├── main.go                    # Main server setup\n├── go.mod                     # Go module definition\n├── config/\n│   └── config.go              # Centralized application configuration\n├── internal/\n│   └── logger/\n│       └── logger.go          # Built-in logger factory\n├── tools/                     # Function-call tool implementations\n│   ├── query_database.go      # Individual tool files (TODO placeholders)\n│   └── send_notification.go\n├── skills/                    # Skill directories (SKILL.md + optional bundled assets)\n│   ├── incident-response/     # Loaded into the system prompt at startup\n│   │   └── SKILL.md\n│   └── support-handoff/\n│       └── SKILL.md\n├── Taskfile.yml               # Development tasks (build, test, lint)\n├── Dockerfile                 # Container configuration\n├── .adl-ignore                # Files to protect from regeneration\n├── .well-known/\n│   └── agent-card.json        # Agent capabilities (auto-generated)\n├── .github/                   # GitHub-specific configurations\n│   ├── workflows/             # Generated when using --ci flag\n│   │   ├── ci.yml             # GitHub Actions CI workflow\n│   │   └── cd.yml             # GitHub Actions CD workflow (with --cd flag)\n│   ├── dependabot.yml         # Generated when scm.dependabot: true\n│   └── ISSUE_TEMPLATE/        # Generated when issue_templates: true\n│       ├── bug_report.md      # Bug report template\n│       ├── feature_request.md # Feature request template\n│       └── refactor_request.md # Refactoring request template\n├── .releaserc.yaml            # Semantic-release configuration (with --cd flag)\n├── k8s/\n│   └── deployment.yaml        # Kubernetes deployment manifest\n├── cloudrun/\n│   └── deploy.sh              # CloudRun deployment script (with --deployment cloudrun)\n├── .flox/                     # Generated when sandbox: flox\n│   ├── env/manifest.toml\n│   ├── env.json\n│   ├── .gitignore\n│   └── .gitattributes\n├── .gitignore                 # Standard Git ignore patterns\n├── .gitattributes             # Git attributes configuration\n├── .editorconfig              # Editor configuration\n├── CLAUDE.md                  # AI assistant instructions (spec.development.ai.claudecode.enabled: true)\n└── README.md                  # Project documentation with setup instructions\n```\n\n### Rust Project Structure\n\n```text\nmy-rust-agent/\n├── src/\n│   ├── main.rs                # Main application entry point\n│   └── tools/                 # Function-call tool implementations\n│       ├── mod.rs             # Module declarations\n│       ├── query_database.rs  # Individual tool implementations\n│       └── send_notification.rs\n├── skills/                    # Skill directories (SKILL.md + optional bundled assets)\n│   ├── incident-response/\n│   │   └── SKILL.md\n│   └── support-handoff/\n│       └── SKILL.md\n├── Cargo.toml                 # Rust package configuration\n├── Taskfile.yml               # Development tasks\n├── Dockerfile                 # Rust-optimized container\n├── .adl-ignore                # Protection configuration\n├── .well-known/\n│   └── agent-card.json        # Agent capabilities\n├── .github/workflows/         # CI configuration (with --ci)\n│   ├── ci.yml                 # Rust-specific CI workflow\n│   └── cd.yml                 # GitHub Actions CD workflow (with --cd flag)\n├── .releaserc.yaml            # Semantic-release configuration (with --cd flag)\n├── k8s/\n│   └── deployment.yaml        # Kubernetes deployment\n├── cloudrun/\n│   └── deploy.sh              # CloudRun deployment script (with --deployment cloudrun)\n├── CLAUDE.md                  # AI assistant instructions (spec.development.ai.claudecode.enabled: true)\n└── README.md                  # Documentation\n```\n\n### Universal Generated Files\n\nAll projects include these essential files regardless of language:\n\n- **`.well-known/agent-card.json`** - A2A agent discovery and capabilities manifest\n- **`Taskfile.yml`** - Unified task runner configuration for build, test, lint, run\n- **`Dockerfile`** - Language-optimized container configuration\n- **`k8s/deployment.yaml`** - Kubernetes deployment manifest\n- **`deploy` task in `Taskfile.yml`** - CloudRun deployment task (when using `--deployment cloudrun`)\n- **`.adl-ignore`** - Protects user implementations from overwrite\n- **CI Workflows** - When using `--ci` flag, generates GitHub Actions workflows:\n  - **GitHub Actions**: `.github/workflows/ci.yml`\n  - **GitLab CI**: `.gitlab-ci.yml` (planned, not yet implemented)\n- **CD Workflows** - When using `--cd` flag, generates continuous deployment:\n  - **GitHub Actions**: `.github/workflows/cd.yml`\n  - **Semantic Release**: `.releaserc.yaml`\n- **Development Environment** - Based on `sandbox` configuration:\n  - **Flox**: `.flox/` directory with environment configuration when `sandbox.flox.enabled: true`\n  - **DevContainer**: `.devcontainer/devcontainer.json` when `sandbox.devcontainer.enabled: true`\n- **AI Assistant Instructions** - Per-agent toggles under `spec.development.ai`\n  (see [Per-agent AI assistants](#per-agent-ai-assistants)):\n  - **CLAUDE.md** when `spec.development.ai.claudecode.enabled: true`\n  - **GEMINI.md** when `spec.development.ai.gemini.enabled: true`\n  - **AGENTS.md** (shared) when any of `codex`, `opencode`, or `infer` is enabled\n\n### CI Integration\n\nWhen using the `--ci` flag, the ADL CLI generates GitHub Actions workflows for your project:\n\n```bash\n# Generate project with CI workflow\nadl generate --file agent.yaml --output ./test-my-agent --ci\n```\n\nThis creates a GitHub Actions workflow (`.github/workflows/ci.yml`) that includes:\n\n- **Automated Testing**: Runs all tests on every push and pull request\n- **Code Quality**: Format checking and linting\n- **Multi-Environment**: Supports main and develop branches\n- **Caching**: Go module caching for faster builds\n- **Task Integration**: Uses the generated Taskfile for consistent build steps\n\nThe generated workflow automatically detects your Go version from the ADL file and configures the appropriate environment.\n\n### CD Integration\n\nThe ADL CLI can generate continuous deployment (CD) pipelines with semantic release automation:\n\n```bash\n# Generate project with CD pipeline\nadl generate --file agent.yaml --output ./test-my-agent --cd\n```\n\nThis creates a complete CD setup including:\n\n- **`.releaserc.yaml`** - Semantic-release configuration with conventional commits\n- **`.github/workflows/cd.yml`** - GitHub Actions CD workflow with manual dispatch\n\nThe generated CD pipeline includes:\n\n- **Semantic Versioning**: Automatic version bumping based on conventional commit messages\n- **Release Automation**: Creates GitHub releases with generated release notes\n- **Container Publishing**: Builds and publishes Docker images to GitHub Container Registry\n- **Multi-Platform Builds**: Supports both AMD64 and ARM64 architectures\n- **Language Detection**: Automatically configures build steps based on your project language\n- **Change Detection**: Only publishes releases when there are changes to release\n\n#### CD Workflow Features\n\n**Manual Trigger**: The CD workflow uses `workflow_dispatch` for controlled releases:\n\n```bash\n# Trigger via GitHub CLI\ngh workflow run cd.yml\n\n# Or trigger via GitHub Actions UI\n```\n\n**Conventional Commits Support**: The pipeline recognizes these commit types for versioning:\n\n- `feat:` - Minor version bump (new features)\n- `fix:` - Patch version bump (bug fixes)\n- `refactor:`, `perf:`, `ci:`, `docs:`, `style:`, `test:`, `build:`, `chore:` - Patch version bump\n\n**Container Registry**: Published images are available at:\n\n```text\nghcr.io/your-org/your-agent:latest\nghcr.io/your-org/your-agent:v1.0.0\nghcr.io/your-org/your-agent:1.0\n```\n\n## CloudRun Deployment\n\nThe ADL CLI provides native support for deploying A2A agents to Google Cloud Run, offering a truly serverless deployment experience without Kubernetes complexity.\n\n### CloudRun Configuration\n\nConfigure CloudRun deployment in your ADL file:\n\n```yaml\nspec:\n  deployment:\n    type: cloudrun\n    cloudrun:\n      image:\n        registry: gcr.io # gcr.io or ghcr.io\n        repository: my-agent # Repository name\n        tag: latest # Image tag\n        useCloudBuild: true # Use Cloud Build or local Docker\n      resources:\n        cpu: \"2\" # CPU allocation (0.1 to 8)\n        memory: 1Gi # Memory limit (128Mi to 32Gi)\n      scaling:\n        minInstances: 0 # Minimum instances (0 to 1000)\n        maxInstances: 100 # Maximum instances (1 to 1000)\n        concurrency: 1000 # Max concurrent requests per instance\n      service:\n        timeout: 3600 # Request timeout in seconds\n        allowUnauthenticated: true # Allow public access\n        serviceAccount: my-agent@PROJECT_ID.iam.gserviceaccount.com\n        executionEnvironment: gen2 # gen1 or gen2\n      environment: # Custom environment variables\n        LOG_LEVEL: info\n        ENVIRONMENT: production\n```\n\n### Container Registry Options\n\n**Google Container Registry (GCR):**\n\n```yaml\nimage:\n  registry: gcr.io\n  repository: my-project/my-agent\n  useCloudBuild: true # Automatically build and push\n```\n\n**GitHub Container Registry (GHCR):**\n\n```yaml\nimage:\n  registry: ghcr.io\n  repository: myorg/my-agent\n  useCloudBuild: false # Skip Cloud Build, use pre-built image\n```\n\n### Generated Deployment Script\n\nWhen using `--deployment cloudrun`, the ADL CLI generates a `deploy` task in the `Taskfile.yml` that:\n\n- **Validates Environment**: Checks for required `PROJECT_ID` and `REGION` variables\n- **Container Building**: Uses Docker locally or Cloud Build based on configuration\n- **Direct gcloud Deployment**: Uses `gcloud run deploy` for serverless deployment\n- **Configuration Summary**: Displays all deployment settings for verification\n\n### CloudRun Deployment Workflow\n\n```bash\n# 1. Generate project with CloudRun deployment\nadl generate --file agent.yaml --output ./my-agent --deployment cloudrun\n\n# 2. Set required environment variables\nexport PROJECT_ID=\"my-gcp-project\"\nexport REGION=\"us-central1\"\n\n# 3. Deploy to CloudRun\ncd my-agent\ntask deploy\n```\n\n### CloudRun with CI/CD\n\nGenerate CloudRun deployment with continuous deployment:\n\n```bash\nadl generate --file agent.yaml --deployment cloudrun --cd\n```\n\nThis creates:\n\n- **CD Workflow**: Automatically deploys to CloudRun after releases\n- **Environment Integration**: Uses GitHub secrets for GCP authentication\n- **Multi-Environment Support**: Deploy to different regions/projects\n\n**Required GitHub Secrets:**\n\n- `GCP_SA_KEY`: Service account key JSON\n- `GCP_PROJECT_ID`: Google Cloud project ID\n- `GCP_REGION`: Deployment region (e.g., us-central1)\n\n### CloudRun Benefits\n\n- **Truly Serverless**: No Kubernetes clusters or infrastructure management\n- **Auto-Scaling**: Scale to zero when idle, scale up automatically under load\n- **Pay-per-Use**: Only pay for actual request processing time\n- **Global Edge**: Deploy to multiple regions with traffic management\n- **Integrated Monitoring**: Built-in logging, metrics, and tracing\n- **Custom Domains**: HTTPS support with automatic SSL certificates\n\n### Example ADL Files\n\nThe CLI includes CloudRun example files:\n\n```bash\n# Validate CloudRun examples\nadl validate examples/cloudrun-agent.yaml\nadl validate examples/cloudrun-ghcr-agent.yaml\n\n# Generate CloudRun projects\nadl generate --file examples/cloudrun-agent.yaml --output ./cloudrun-test\nadl generate --file examples/cloudrun-ghcr-agent.yaml --output ./ghcr-test\n```\n\n## Sandbox Environments\n\nThe ADL CLI supports multiple development environments for isolated, reproducible development:\n\n### Flox Environment\n\nConfigure Flox for your project by adding to your ADL file:\n\n```yaml\nspec:\n  development:\n    sandbox:\n      flox:\n        enabled: true\n```\n\nGenerated files:\n\n- `.flox/env/manifest.toml` - Flox environment manifest with language-specific services\n- `.flox/env.json` - Environment configuration\n- `.flox/.gitignore` - Flox-specific ignore patterns\n- `.flox/.gitattributes` - Git attributes for Flox files\n\n### DevContainer Environment\n\nConfigure DevContainer for your project:\n\n```yaml\nspec:\n  development:\n    sandbox:\n      devcontainer:\n        enabled: true\n```\n\nGenerated files:\n\n- `.devcontainer/devcontainer.json` - VS Code DevContainer configuration with language support\n\n### Multiple Environment Support\n\nYou can enable multiple sandbox environments simultaneously:\n\n```yaml\nspec:\n  development:\n    sandbox:\n      flox:\n        enabled: true\n      devcontainer:\n        enabled: true\n```\n\nThis generates both Flox and DevContainer configurations, allowing developers to choose their preferred environment.\n\n### Benefits of Sandbox Environments\n\n- **Reproducible Development** - Consistent environments across team members\n- **Isolated Dependencies** - No conflicts with system-wide installations\n- **Language-Specific Tooling** - Pre-configured with appropriate development tools\n- **CI/CD Integration** - Matches production environment characteristics\n\n## Enterprise Features\n\n### Authentication Configuration\n\nEnable server authentication in your ADL file:\n\n```yaml\nspec:\n  server:\n    port: 8443\n    debug: false\n    auth:\n      enabled: true\n```\n\nThis generates enterprise-ready authentication scaffolding in your project.\n\n### SCM Integration\n\nConfigure source control management for automatic CI/CD provider detection:\n\n```yaml\nspec:\n  scm:\n    provider: github # gitlab support planned\n    url: \"https://github.com/company/my-agent\"\n    github_app: false # optional: enable GitHub App for CD\n    issue_templates: true # optional: generate GitHub issue templates\n    dependabot: true # optional: generate Dependabot configuration\n```\n\n**Features:**\n\n- **Automatic CI Detection** - Generates appropriate workflows based on SCM provider\n- **Repository Integration** - Links generated projects to source control\n- **Workflow Optimization** - SCM-specific optimizations and best practices\n- **GitHub App Support** - Enhanced security for enterprise CD pipelines\n- **Issue Templates** - Generate GitHub issue templates for standardized bug reports and feature requests\n- **Dependabot Configuration** - Auto-generate `.github/dependabot.yml` for dependency upgrades\n\n#### GitHub App Integration\n\nFor enterprise environments, you can enable GitHub App-based CD deployment for enhanced security:\n\n```yaml\nspec:\n  scm:\n    provider: github\n    url: \"https://github.com/company/my-agent\"\n    github_app: true\n```\n\n**GitHub App CD Benefits:**\n\n- **Enhanced Security** - App tokens are automatically revoked after pipeline execution\n- **Enterprise Compliance** - Keeps main branch protected from direct pushes\n- **Bot Identity** - Release operations performed by dedicated bot account\n- **Audit Trail** - Clear attribution of automated actions\n\n**Required GitHub Secrets:**\n\n- `BOT_GH_APP_ID` - Your GitHub App ID\n- `BOT_GH_APP_PRIVATE_KEY` - Your GitHub App private key\n\nWhen `github_app: true` is set, the generated CD pipeline will use GitHub App authentication instead of the default `GITHUB_TOKEN`, providing better security isolation for release management.\n\n### AI Provider Support\n\nThe ADL CLI supports multiple AI providers including OpenAI, Anthropic, DeepSeek, Ollama (for local LLMs), Google AI, Mistral, and Groq. Each provider requires appropriate API keys to be configured as environment variables. See the ADL examples above for configuration details.\n\n## Artifacts Support\n\nEnable artifacts support to allow your agent to create, store, and manage files and resources:\n\n```yaml\nspec:\n  artifacts:\n    enabled: true\n```\n\nConfigure storage via environment variables (see generated README for A2A_ARTIFACT_* variables). Supports both filesystem and MinIO/S3 storage backends.\n\n**Examples:**\n- `examples/go-agent-artifacts-filesystem.yaml` - Filesystem storage example\n- `examples/go-agent-artifacts-minio.yaml` - MinIO storage example\n\n## GitHub Issue Templates\n\nThe ADL CLI can automatically generate GitHub issue templates for your agent projects, providing standardized forms for bug reports, feature requests, and refactoring tasks:\n\n```yaml\nspec:\n  scm:\n    provider: github\n    url: \"https://github.com/company/my-agent\"\n    issue_templates: true # Enable issue template generation\n```\n\nWhen `issue_templates: true` is set, the following templates are generated in `.github/ISSUE_TEMPLATE/`:\n\n- **`bug_report.md`** - Structured bug reporting with severity levels, reproduction steps, and environment details\n- **`feature_request.md`** - Feature proposals with use case descriptions and acceptance criteria\n- **`refactor_request.md`** - Code improvement requests with motivation and impact analysis\n\n**Issue Template Features:**\n\n- **Agent Context** - Templates include agent name and version from your ADL metadata\n- **Structured Sections** - Consistent formatting for better issue triage and tracking\n- **GitHub Integration** - Automatic labels and assignees configured in frontmatter\n- **Severity Levels** - Priority classification for bug reports (critical, high, medium, low)\n- **Environment Info** - Sections for capturing logs, system details, and configurations\n\n## Dependabot Configuration\n\nThe ADL CLI can generate a `.github/dependabot.yml` manifest so generated agent\nprojects keep their dependencies up to date automatically. The feature defaults\nto `false` and is opted into via `spec.scm.dependabot`:\n\n```yaml\nspec:\n  scm:\n    provider: github\n    url: \"https://github.com/company/my-agent\"\n    dependabot: true # Enable Dependabot configuration\n```\n\nWhen `dependabot: true` is set (and the SCM provider is GitHub), the generator\nemits a weekly-schedule manifest covering the ecosystems present in your ADL:\n\n- **Language ecosystem** - `gomod`, `cargo`, or `npm` (selected from `spec.language`)\n- **`github-actions`** - Keeps `.github/workflows/` actions pinned\n- **`docker`** - Tracks the base image in the generated `Dockerfile`\n- **`devcontainers`** - Included when `spec.development.sandbox.devcontainer.enabled: true`\n\nEach ecosystem groups all updates into a single PR per week to keep noise low.\nThe default of `false` keeps the existing behavior unchanged for projects that\nmanage dependency upgrades themselves.\n\n## Examples\n\nThe CLI includes example ADL files in the `examples/` directory:\n\n```bash\n# Validate examples\nadl validate examples/go-agent.yaml\nadl validate examples/rust-agent.yaml\nadl validate examples/github-app-agent.yaml\nadl validate examples/cloudrun-agent.yaml\nadl validate examples/cloudrun-ghcr-agent.yaml\n\n# Generate from examples\nadl generate --file examples/go-agent.yaml --output ./test-go-agent\nadl generate --file examples/rust-agent.yaml --output ./test-rust-agent\nadl generate --file examples/github-app-agent.yaml --output ./test-github-app-agent --cd\nadl generate --file examples/cloudrun-agent.yaml --output ./test-cloudrun-agent --deployment cloudrun\nadl generate --file examples/cloudrun-ghcr-agent.yaml --output ./test-ghcr-agent --deployment cloudrun\n\n# Generate with CI/CD pipeline\nadl generate --file examples/github-app-agent.yaml --output ./enterprise-agent --ci --cd\nadl generate --file examples/cloudrun-agent.yaml --output ./cloudrun-enterprise --deployment cloudrun --cd\n```\n\n**Example ADL Files:**\n\n- `go-agent.yaml` - Basic Go agent with multiple skills and capabilities\n- `rust-agent.yaml` - Rust agent with enterprise features\n- `github-app-agent.yaml` - Enterprise agent with GitHub App CD integration\n- `cloudrun-agent.yaml` - CloudRun deployment with Google Container Registry\n- `cloudrun-ghcr-agent.yaml` - CloudRun deployment with GitHub Container Registry\n\n## Template System \u0026 Architecture\n\nThe ADL CLI uses a sophisticated template system that generates language-specific projects:\n\n### Language Detection\n\nThe generator automatically detects your target language from the ADL file:\n\n```go\n// Automatic detection based on spec.language configuration\nfunc DetectLanguageFromADL(adl *schema.ADL) string {\n    if adl.Spec.Language.Go != nil     { return \"go\" }\n    if adl.Spec.Language.Rust != nil   { return \"rust\" }\n    if adl.Spec.Language.TypeScript != nil { return \"typescript\" }\n    return \"go\" // default\n}\n```\n\n### File Mapping System\n\nEach language has its own file mapping that determines what gets generated:\n\n**Go Projects:**\n\n- `main.go` → Go main server setup\n- `tools/{toolname}.go` → Individual function-call tool implementations\n- `skills/{skillid}/SKILL.md` → Markdown skill playbooks (loaded into system prompt at runtime)\n- `go.mod` → Go module configuration\n- Language-specific Dockerfile and CI configurations\n\n**Rust Projects:**\n\n- `src/main.rs` → Rust main application\n- `src/tools/{toolname}.rs` → Tool descriptor + handler\n- `src/tools/mod.rs` → Module declarations\n- `skills/{skillid}/SKILL.md` → Markdown skill playbooks\n- `Cargo.toml` → Rust package configuration\n\n**Universal Files:**\n\n- `Taskfile.yml` → Development task runner\n- `.well-known/agent-card.json` → A2A capabilities manifest (skills are listed here, not tools)\n- `k8s/deployment.yaml` → Kubernetes deployment\n- CI workflows and sandbox configurations\n\n### Template Context\n\nAll templates receive a rich context object:\n\n```go\ntype Context struct {\n    ADL      *schema.ADL           // Complete ADL configuration\n    Metadata GeneratedMetadata     // Generation metadata\n    Language string               // Detected language\n}\n```\n\nThis allows templates to access any ADL configuration and generate language-appropriate code.\n\n## Customizing Generation with .adl-ignore\n\nThe ADL CLI automatically creates a `.adl-ignore` file during project generation to protect files containing TODO implementations. This file works similar to `.gitignore` and prevents important implementation files from being overwritten during subsequent generations.\n\n### Automatically Protected Files\n\nWhen you generate a project, implementation files are automatically added to `.adl-ignore` to protect your business logic from being overwritten during regeneration.\n\nYou can control which additional files are generated or updated by editing the `.adl-ignore` file:\n\n```bash\n# .adl-ignore\n# Skip Docker-related files if you have custom containerization\nDockerfile\ndocker-compose.yml\n\n# Skip Kubernetes manifests if you use different deployment tools\nk8s/\n\n# Skip specific generated files you want to customize\nmiddleware.go\nauth.go\n\n# Skip build configuration if you have custom setup\nTaskfile.yml\n```\n\n### .adl-ignore Patterns\n\n- Use `#` for comments\n- Use `/` at the end to match directories\n- Use `*` for wildcards\n- Exact file paths or glob patterns\n- Protects files during all `generate` operations\n\n### Common Use Cases\n\n- **Custom Deployment**: Skip `Dockerfile`, `k8s/`, `docker-compose.yml`\n- **Custom Build**: Skip `Taskfile.yml`, `Makefile`\n- **Custom Auth**: Skip `auth.go`, `middleware.go`\n- **Custom Documentation**: Skip `README.md`\n\n## Development\n\n### Prerequisites\n\n- Go 1.26.2+\n- [Task](https://taskfile.dev/) (optional, for using Taskfile commands)\n\n### Building from Source\n\n```bash\ngit clone https://github.com/inference-gateway/adl-cli.git\ncd adl-cli\n\n# Install dependencies\ngo mod download\n\n# Build\ntask build\n\n# Run tests\ntask test\n\n# Format code\ntask fmt\n\n# Lint\ntask lint\n```\n\n### Testing\n\n```bash\n# Run tests\ntask test\n\n# Test with coverage\ntask test:coverage\n\n# Test all examples\ntask examples:test\n\n# Generate all examples\ntask examples:generate\n```\n\n### Contributing\n\n1. Fork the repository\n2. Create a feature branch\n3. Make your changes\n4. Add tests for new functionality\n5. Run `task ci` to ensure everything passes\n6. Submit a pull request\n\n## Configurable Acronyms\n\nThe ADL CLI includes support for configurable acronyms to improve code generation readability. This feature helps generate more readable function and struct names by properly capitalizing acronyms in generated code.\n\n### How It Works\n\nDefine custom acronyms in your ADL file's `spec.acronyms` field. These acronyms will be properly capitalized when generating identifiers in your code.\n\n### Configuration\n\n```yaml\nspec:\n  language:\n    go:\n      module: \"github.com/company/my-agent\"\n      version: \"1.26.2\"\n  acronyms: [\"n8n\", \"xml\", \"mqtt\", \"iot\", \"uuid\"]\n```\n\n### Generated Code Examples\n\n**Without custom acronyms:**\n\n- `get_n8n_docs` → `GetN8nDocsSkill`\n- `process_xml_data` → `ProcessXmlDataSkill`\n\n**With custom acronyms:**\n\n- `get_n8n_docs` → `GetN8NDocsSkill`\n- `process_xml_data` → `ProcessXMLDataSkill`\n\n### Default Acronyms\n\nThe following acronyms are recognized by default:\n\n- **Common**: id, api, url, uri, json, xml, sql, html, css, js, ui, uuid\n- **Network**: http, https, tcp, udp, ip, dns, tls, ssl\n- **Tech**: cpu, gpu, ram, io, os, db\n\nYour custom acronyms extend these defaults and take precedence over them.\n\n## Post-Generation Hooks\n\nThe ADL CLI supports custom post-generation hooks that run automatically after project generation. These hooks allow you to execute commands like formatting, linting, testing, or custom setup scripts.\n\n### Default Hooks\n\nEach language has sensible defaults:\n\n**Go Projects:**\n\n- `go fmt ./...` - Format all Go source files\n- `go mod tidy` - Download dependencies and clean up go.mod\n\n**Rust Projects:**\n\n- `cargo fmt` - Format all Rust source files\n- `cargo check` - Check the project for errors\n\n### Custom Hooks\n\nYou can customize or extend the default behavior by adding a `hooks` section to your ADL file:\n\n```yaml\napiVersion: adl.inference-gateway.com/v1\nkind: Agent\nmetadata:\n  name: my-agent\nspec:\n  # ... other configuration ...\n\n  # Custom post-generation hooks\n  hooks:\n    post:\n      - \"go fmt ./...\"\n      - \"go mod tidy\"\n      - \"go vet ./...\"\n      - \"go test -short ./...\"\n      - \"golangci-lint run --fix\"\n```\n\n### Hooks Behavior\n\n- **Override Defaults**: When you specify custom hooks, they completely replace the language defaults\n- **Command Execution**: Commands run in the generated project directory\n- **Error Handling**: Failed commands show warnings but don't stop generation\n- **Sequential Execution**: Commands run in the order specified\n- **Shell Support**: Commands are executed through the system shell\n\n### Example Configurations\n\n**Extended Go Development:**\n\n```yaml\nhooks:\n  post:\n    - \"go mod download\" # Download dependencies first\n    - \"go generate ./...\" # Generate code if needed\n    - \"gofumpt -l -w .\" # Improved formatting\n    - \"golangci-lint run --fix\" # Lint and auto-fix\n    - \"go test -race -short ./...\" # Run tests\n    - \"go build -v ./...\" # Verify build works\n```\n\n**Rust with Additional Tools:**\n\n```yaml\nhooks:\n  post:\n    - \"cargo fmt\"\n    - \"cargo clippy --fix --allow-dirty\"\n    - \"cargo check --all-targets\"\n    - \"cargo test --lib\"\n```\n\n**TypeScript/Node.js:**\n\n```yaml\nhooks:\n  post:\n    - \"npm install\"\n    - \"npm run format\"\n    - \"npm run lint:fix\"\n    - \"npm run type-check\"\n    - \"npm test\"\n```\n\n### Best Practices\n\n- **Keep hooks fast** - Avoid long-running commands that slow down generation\n- **Use error-tolerant commands** - Commands should gracefully handle missing tools\n- **Order matters** - Place dependencies first (e.g., `npm install` before `npm run lint`)\n- **Document requirements** - Note any required tools in your project README\n\n## Roadmap\n\n### Language Support\n\nThe ADL CLI currently supports Go and Rust, with plans to expand to additional programming languages:\n\n#### ✅ Currently Supported\n\n- **Go** - Full support with templates for main.go, go.mod, and tools\n- **Rust** - Full support with templates for main.rs, Cargo.toml, and tools\n\n#### 🚧 Planned Support\n\n- **TypeScript/Node.js** - Template structure exists but templates not yet implemented\n  - Complete A2A agent generation with Express.js framework planned\n  - AI-powered agents with OpenAI/Anthropic integration\n  - Enterprise features (auth, metrics, logging)\n  - Docker and Kubernetes deployment configs\n\n- **Python** - Rapid prototyping and AI-first development\n  - FastAPI-based server generation\n  - Rich AI ecosystem integration\n  - Jupyter notebook support for development\n\n#### 🔮 Future Considerations\n\n- **Java/Kotlin** - Enterprise JVM support\n- **C#/.NET** - Microsoft ecosystem integration\n- **Swift** - Apple ecosystem and server-side Swift\n\n### Template Enhancements\n\n- **Multi-language projects** - Generate polyglot agents with language-specific microservices\n- **Custom templates** - User-defined project templates and scaffolding\n- **Plugin system** - Extensible architecture for custom generators\n- **Cloud-native templates** - Serverless (AWS Lambda, Vercel) and edge deployment support\n\n### Contribute to the Roadmap\n\nWe welcome community input on our roadmap! Please:\n\n- 💡 Suggest new languages or frameworks via [Issues](https://github.com/inference-gateway/adl-cli/issues)\n- 🤝 Contribute implementations for new languages (see [Contributing Guide](CONTRIBUTING.md))\n\n## License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## Support\n\n- 📖 [Documentation](https://docs.inference-gateway.com)\n- 💬 [Discussions](https://github.com/inference-gateway/adl-cli/discussions)\n- 🐛 [Issues](https://github.com/inference-gateway/adl-cli/issues)\n\n---\n\n\u003e 🤖 Powered by the [Inference Gateway ecosystem](https://github.com/inference-gateway/)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Finference-gateway%2Fadl-cli","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Finference-gateway%2Fadl-cli","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Finference-gateway%2Fadl-cli/lists"}