Brooke

This commit performs a critical **dependency upgrade** by updating the **Node.js runtime version** to 22 across the project. This **security-driven maintenance** addresses a recent vulnerability, ensuring the **CLI** and all related development and CI/CD environments operate on a patched and secure foundation. The change, documented in a new changeset file, primarily impacts **developer setups** and **automated build processes** by updating the `.nvmrc` file.

This commit introduces a **non-interactive mode** for the **CLI `deploy` command**, enabling automated deployments without requiring user intervention. The core implementation in `packages/cli/src/commands/deploy/index.ts` modifies existing handlers like `handleInitDeployment` to adjust JSON output and error handling when the client's `nonInteractive` flag is active. A new utility function, `getCommandNameWithGlobalFlags`, was added to `packages/cli/src/util/arg-common.ts` to support command suggestions in this mode. This **new feature** significantly improves the `deploy` command's suitability for **CI/CD pipelines** and scripting by allowing programmatic execution.

This commit introduces a **new feature** to the **CLI** by adding a **non-interactive mode** for the **build command**. This enhancement allows users to execute `build` operations without manual prompts, making it suitable for automated environments and scripts. The change affects the **CLI's build process** and its **agent output constants**, specifically by adding a `PROJECT_SETTINGS_REQUIRED` constant to handle scenarios where project settings are missing in non-interactive mode. This improves the **CLI's automation capabilities** and provides clearer feedback for script-based usage.

This commit **fixes a regression** in the **CLI's login command** that caused non-interactive login attempts to prematurely exit. Previously, the system would incorrectly terminate the login flow with an `action_required` message in non-interactive environments. The change **removes the problematic non-interactive login logic**, ensuring the login process now correctly continues as intended. This **restores proper functionality for automated or script-based login scenarios** within the CLI.

This commit significantly enhances the **reliability** of the project's **CI/CD pipeline** by introducing **retry mechanisms** for automated tests. The `test-e2e.yml` and `test.yml` GitHub Actions workflows are updated to automatically re-run failed tests and enforce a maximum parallelism limit, directly addressing issues with flaky test failures. This **maintenance** effort aims to reduce false negatives in CI, providing more consistent and trustworthy feedback to developers. The changes are documented in a new changeset file, highlighting the improvements to test stability.

This commit introduces a **non-interactive mode** for the **CLI's webhook commands**, significantly enhancing their utility for automation and scripting. It enables `webhooks create`, `get`, `ls`, and `rm` commands within the `@vercel/cli` package to produce **structured JSON output** for both successful operations and detailed error messages, including suggested next actions. This **new feature** allows for programmatic management of webhooks, providing consistent, machine-readable responses. New agent output constants were added to support this, and comprehensive unit tests ensure the reliability of these non-interactive modes across various scenarios.

This commit introduces a **new capability** to the **CLI's `flags` commands**, enabling a **non-interactive mode** for operations such as `flags disable`, `flags enable`, `flags sdk-keys add`, and `flags sdk-keys rm`. It implements robust non-interactive error handling, command suggestions, and confirmations across these specific commands, significantly enhancing their **scriptability and automation**. New utility functions like `getGlobalFlagsFromArgv` and `buildCommandWithGlobalFlags` were added to support this mode, alongside comprehensive unit tests for the affected commands. This change improves the **developer experience** by allowing programmatic management of feature flags without requiring user interaction.

This commit introduces a **non-interactive mode** across all **CLI `routes` commands**, enabling automated execution without user prompts. It enhances the `routes add` command with **AI-powered route generation** in this mode, alongside robust error handling and agent output for all `routes` subcommands like `delete`, `edit`, and `publish`. This **new feature** significantly improves the **scriptability and automation capabilities** for managing routes, providing clearer feedback for programmatic interactions. The changes span core `routes` command logic, shared utilities in `packages/cli/src/commands/routes/shared.ts`, and agent output mechanisms, ensuring consistent behavior and better integration into automated workflows.

This commit introduces a **new capability** by implementing a **non-interactive mode** for several **CLI `redirects` commands**. This enhancement allows for automated execution of `add`, `promote`, `remove`, `restore`, and `upload` subcommands, providing structured **JSON output** for success, errors, and required actions, and often requiring a `--yes` flag for confirmation. It includes new utility functions in `packages/cli/src/commands/redirects/shared.ts` for robust argument parsing and error handling, alongside new agent output constants for machine-readable responses. This significantly enhances the **CLI's automation capabilities**, enabling `redirects` management to be seamlessly integrated into **CI/CD pipelines and scripting workflows**.

This commit **adds comprehensive documentation** detailing best practices for implementing and consuming the **CLI in non-interactive mode**. A new document, `packages/cli/docs/non-interactive-mode.md`, is introduced to guide users and developers on this topic. Additionally, a cross-reference to this new guide is added within `packages/cli/AGENTS.md` to improve discoverability. This **documentation addition** enhances the clarity and usability of the CLI for automated scripts and environments, serving as a **maintenance** and **feature enhancement** for the project's user experience.

This commit introduces a **non-interactive mode** for several **`dns` and `domains` CLI commands**, significantly enhancing their **automation and scriptability** for use in CI/CD pipelines or programmatic interactions. It implements non-interactive error handling and action prompts across `dns add`, `dns import`, `dns ls`, `dns rm`, and `domains add`, while explicitly preventing non-interactive execution for `domains buy`. This **new capability** involves updating the `packages/cli` module, including new constants in `agent-output-constants.ts` for standardized output and refined error reporting in `add-domain-to-project.ts`. Unit tests have been added to ensure the correct behavior of these non-interactive modes, particularly for confirmation requirements.

This commit introduces a **non-interactive mode** for several **`teams` CLI commands**, enabling programmatic execution without user prompts. This **new capability** specifically allows for non-interactive usage of **`teams add`**, **`teams invite`**, and **`teams switch`**, while also enhancing error handling for **`teams list`**. The changes include adding new CLI options like `--slug` and `--name` for `teams add`, implementing robust argument validation, and improving agent-specific error reporting for automated workflows. This significantly enhances the CLI's utility for **automation and scripting** of team management tasks, supported by new utility functions for argument parsing and comprehensive unit tests.

This commit introduces a **new feature** by adding **non-interactive mode support** for the **CLI's `rolling-release` commands**, enabling automated deployments without user prompts. It implements comprehensive **error handling** for these non-interactive scenarios, guiding users on required flags like `--yes` for non-production deployments, and includes new helper functions such as `buildDeploymentSuggestionCommands`. The changes primarily affect the `packages/cli/src/commands/rolling-release` module, updating help texts and adding dedicated unit tests for the `configure` command's non-interactive behavior. This enhancement significantly improves the utility of rolling releases for CI/CD pipelines and scripting environments.

This commit introduces a **new feature** to the **CLI `mcp` command**, enabling a **non-interactive mode** for automated execution. Users can now specify target clients directly using the new `--clients` flag, bypassing interactive prompts and receiving output in a machine-readable JSON format. This enhancement primarily affects the `packages/cli/src/commands/mcp` module, with updates to `command.ts`, `constants.ts`, `index.ts`, and `mcp.ts` to parse the new flag, validate client selections, and adjust output. Additionally, `packages/cli/src/util/agent-output.ts` was refactored to ensure consistent JSON output, significantly improving the command's utility for scripting and integration workflows.

This commit introduces a **new feature** to the **Vercel CLI**, enabling a **non-interactive mode** for the `login` command. When `vercel login` is executed without arguments in a non-interactive environment, it will now output an `action_required` message, providing clearer feedback to the user. This enhancement improves the usability of the `login` command in automated scripts or CI/CD pipelines, as it prevents the command from hanging and instead informs about necessary actions. The change primarily affects the **CLI `login` command's output behavior** and includes new unit tests to verify this functionality.

This commit significantly **enhances the CLI evaluation framework** by introducing new authentication-related tests and improving the underlying testing infrastructure. It adds **new evaluation tests** for `login-whoami` and `login-not-logged-in` scenarios, expanding coverage for critical authentication flows within the `packages/cli/evals` module. Concurrently, existing `env/add` and `env/update` evaluation scripts are **refactored** to remove project-state assertions, improving their reliability. Furthermore, the evaluation runner is **upgraded to support an `auth-state` matrix**, enabling tests to be executed across various authentication states. These changes are purely **test infrastructure improvements**, providing better coverage and robustness for the CLI without affecting production code.

This commit **refactors** the **CLI evaluation tests** to improve how shell commands are read. Previously, tests relied on `command-used.txt` files, but now they **parse `results.json`** for command retrieval. This **test refactoring** impacts various `env` and `non-interactive` CLI tests within `packages/cli/evals`, standardizing the process with a new `getShellCommands` helper. The change **enhances the maintainability and reliability of the CLI test suite** by updating test logic and associated `PROMPT.md` documentation, without affecting any production code or user-facing functionality.

This commit **enhances the CLI evaluation infrastructure** by introducing a **skills matrix dimension**, enabling tests to be executed both with and without a `skills` directory present. This **test infrastructure enhancement** involves adding new functions like `getSkillsVariantsFromEnv` and `getEvalVariants` within `packages/cli/evals` to manage these test variants, and configuring the **GitHub Actions workflow** (`cli-evals.yml`) to leverage this new capability. The change ensures more comprehensive testing of the CLI by simulating different project setups, with the behavior controlled by the new `CLI_EVAL_SKILLS_MODES` environment variable, and has **no impact on production code or user-facing features**.

This commit introduces **matrix support for CLI evals** by implementing **setup and destroy hooks** within the `cli/evals` subsystem. It adds new `hooks.ts` to manage **ephemeral project creation** and teardown based on different project modes, and **refactors `run.ts`** to integrate these hooks and support project mode variants. This **enhances the testing infrastructure** for CLI commands, allowing for more comprehensive and isolated evaluation scenarios. The work primarily involves **new capabilities** and **refactoring** of test-related code, including updates to the CI workflow (`cli-evals.yml`) for increased timeout and memory limits, without affecting production code.

This commit introduces a **new capability** by adding an **hourly cron schedule** to the existing **GitHub Actions workflow** responsible for running **CLI evaluations**. This enhancement automates the periodic execution of these evaluations, ensuring more frequent and consistent testing of the command-line interface. The change primarily affects the **CI/CD pipeline** by modifying `.github/workflows/cli-evals.yml` to include the cron trigger, and also adds documentation via `.changeset/cli-evals-hourly-cron.md`. This **maintenance automation** aims to improve the **stability and quality** of the CLI by enabling earlier detection of potential issues, without altering any core business logic.