Claude Code 2.1.88

repository·main·Indexed 41 days ago

https://github.com/exhen/claude-code-2.1.88

An agentic terminal-based coding tool by Anthropic that uses natural language to execute routine tasks, explain code, and manage git workflows. It supports Model Context Protocol (MCP) server management, custom auto-mode classifier rules, and authentication via Anthropic Console or claude.ai.

Tokens
16.6K
Snippets
41
Records
83
Agent score
68%

What's inside Claude Code

  1. Install Claude Code

    main

    To use Claude Code as an agentic coding tool in your terminal, install it globally using npm. Once installed, you can run the claude command from within any project directory to start an interactive session.

    npm install -g @anthropic-ai/claude-code
  2. Untitled record

    main

    The provided file is a compiled/minified React component (ManageMarketplaces.tsx) containing encoded data (likely a source map or a serialized component structure). It does not contain human-readable CLI commands, flags, or usage instructions for an end-user.

    Because the content is a serialized binary/encoded format (TSX/Source Map segment), no actionable CLI or API knowledge can be extracted for a developer looking to use the ManageMarketplaces feature.

  3. Understand Ultrareview billing and quotas

    main

    Ultrareview usage is subject to quota and billing rules based on your subscription plan:

    • Team and Enterprise Plans: Include ultrareview; no free-review quota or extra usage dialogs are shown.
    • Consumer Plans (Pro/Max):
      • You have a set number of free reviews.
      • Once free reviews are exhausted, the tool checks if Extra Usage is enabled in your settings.
      • If Extra Usage is enabled and you have a sufficient balance, the review will proceed and bill as Extra Usage.
      • If Extra Usage is not enabled or your balance is too low (less than 10 credits), the review cannot launch.

    When a review is launched using Extra Usage, the tool will append a billing note to the launch notification (e.g., This review bills as Extra Usage.).

  4. Thinkback Skill Modes

    main

    When interacting with the thinkback skill via Claude, you can use specific modes to manage your year-in-review animation. These modes are triggered by specific prompts that instruct Claude to use the Skill tool:

    ModePurposePrompt/Instruction
    editModify existing contentUse the Skill tool to invoke the "thinkback" skill with mode=edit to modify my existing Claude Code year in review animation...
    fixResolve errorsUse the Skill tool to invoke the "thinkback" skill with mode=fix to fix validation or rendering errors...
    regenerateStart freshUse the Skill tool to invoke the "thinkback" skill with mode=regenerate to create a completely new Claude Code year in review animation from scratch...
  5. Install a plugin from the Marketplace

    main

    When interacting with the Marketplace UI, you can trigger installations through several actions:

    • User Scope: Use the install-user action to install a plugin for the current user.
    • Project Scope: Use the install-project action to install a plugin for the current project.
    • Local Scope: Use the install-local action to install a plugin locally.

    If an installation fails, the system provides feedback via the error state, which includes the error message and details. If the installation is successful, the system provides a success message.

    Note: For plugins that are already globally installed, the system will skip the installation step and notify you that it is already installed.

  6. Untitled record

    main

    The BrowseMarketplace component is a UI interface within the Claude Code CLI that allows users to explore available plugins. It displays a list of marketplace plugins, their names, and metadata such as version, description, and author.

    Key features of the marketplace view include:

    • Plugin Details: Shows the plugin name, version, description, and total number of plugins installed.
    • Actionable Commands: If a plugin is already installed, the UI provides specific commands to manage it (e.g., command: <command_name>).
    • Installation Status: Indicates whether a plugin is currently installed or if it is available for installation.
    • Error Handling: Displays installation errors if a command fails.
    • Dynamic Loading: Supports scrolling and loading more items as the user navigates the list.

    Note: The component uses a plugin-details view mode to show specific metadata for a selected plugin, including its source (e.g., github, npm, pip, or url) and available hooks/commands.

  7. Handle GitHub App OAuth flow in Claude Code

    main

    When installing the GitHub App, Claude Code uses an OAuth flow to create a long-lived authentication token for GitHub Actions. The process follows these states:

    1. Starting: The initial state where the token creation process begins.
    2. Waiting for login: The CLI provides a URL for signing in. If the browser does not open automatically, you can copy the URL (shortcut c) and paste it into your browser.
    3. Paste code prompt: After signing in via the browser, you may be prompted to paste an authentication code back into the CLI.
    4. Processing: The CLI is verifying the authentication.
    5. Success: The authentication token is created successfully and will be used for GitHub Actions setup.
    6. Error: If an error occurs, the CLI displays the error message. If the error is retryable, you can press Enter to try again or any other key to cancel. Otherwise, press any key to return to API key selection.
    7. About to retry: A transient state shown when the CLI is attempting to retry the authentication.
  8. How XAA (SEP-990) works with MCP

    main

    XAA (SEP-990) can be enabled for an MCP server using the --xaa flag during the add command.

    Requirements:

    1. Environment: You must have CLAUDE_CODE_ENABLE_XAA=1 set in your environment.
    2. Setup: You must have previously run claude mcp xaa setup to configure settings.xaaIdp.
    3. Authentication: You must provide both --client-id and either --client-secret or have the secret available via environment variables.

    Note: XAA is only supported for http and sse transports. It will be ignored if using stdio transport.

  9. Plugin installation scopes

    main

    When installing a plugin, you can specify the scope of the installation. This determines where the plugin is available:

    • user: Installed at the user level, making it available globally across all Claude Code instances for that user.
    • project: Installed specifically for the current project context.
    • local: Installed in a local scope.

    If a plugin requires configuration immediately after installation, the UI will automatically transition to a PluginOptionsFlow to guide you through the setup.

  10. Plugin Metadata and Components

    main

    When browsing plugins, the UI displays several key pieces of metadata to help you decide what to install:

    • Identity: Name, version, and author.
    • Description: A brief summary of the plugin's purpose.
    • Capabilities (Components): A summary of what the plugin adds to Claude Code:
      • Commands: CLI commands the plugin provides.
      • Agents: Specialized AI agents.
      • Hooks: Lifecycle or event hooks.
      • MCP Servers: Model Context Protocol servers.
    • Popularity: For the official marketplace, the number of existing installs is displayed.
  11. Manage plugin installation scopes

    main

    Claude Code supports different installation scopes for plugins, allowing you to manage how plugins are applied to your environment:

    • User Scope (install:user): Installs the plugin for the current user. This is generally the preferred method for most plugins.
    • Project Scope (install-project): Installs the plugin specifically for the current project context.
    • Local Scope (install-local): Installs the plugin locally.

    When performing an installation, the system uses the installPluginFromMarketplace utility to handle the logic based on the selected scope.