task-master-ai@0.43.0

Minor Changes

• #1599 e689fcf Thanks @triepod-ai! - Add MCPB bundle for single-click Claude Desktop installation

  • Added manifest.json for MCP Bundle (MCPB) specification v0.3
  • Added .mcpbignore to exclude development files from bundle
  • Added icon.png (512x512) for Claude Desktop display
  • Enables users to install Task Master MCP server directly in Claude Desktop without manual configuration

• #1605 efedc85 Thanks @Crunchyman-ralph! - Add verbose output mode to loop command with --verbose flag

  • New -v, --verbose flag shows Claude's work in real-time (thinking, tool calls) rather than waiting until the iteration completes
  • New --no-output flag excludes full Claude output from iteration results to save memory
  • Improved error handling with proper validation for incompatible options (verbose + sandbox)

• #1611 c798639 Thanks @Crunchyman-ralph! - Add optional metadata field to tasks for storing user-defined custom data

  Tasks and subtasks now support an optional metadata field that allows storing arbitrary JSON data such as:

  • External IDs (GitHub issues, Jira tickets, Linear issues)
  • Workflow data (sprints, story points, custom statuses)
  • Integration data (sync timestamps, external system references)
  • Custom tracking (UUIDs, version numbers, audit information)

  Key features:

  • AI-Safe: Metadata is preserved through all AI operations (update-task, expand, etc.) because AI schemas intentionally exclude this field
  • Flexible Schema: Store any JSON-serializable data without schema changes
  • Backward Compatible: The field is optional; existing tasks work without modification
  • Subtask Support: Both tasks and subtasks can have their own metadata
  • MCP Tool Support: Use update_task and update_subtask with the metadata parameter to update metadata (requires TASK_MASTER_ALLOW_METADATA_UPDATES=true in MCP server environment)

  Example usage:

  {
    "id": 1,
    "title": "Implement authentication",
    "metadata": {
      "githubIssue": 42,
      "sprint": "Q1-S3",
      "storyPoints": 5
    }
  }

  MCP metadata update example:

  // With TASK_MASTER_ALLOW_METADATA_UPDATES=true set in MCP env
  update_task({
    id: "1",
    metadata: '{"githubIssue": 42, "sprint": "Q1-S3"}',
  });

Patch Changes

• #1587 0d628ca Thanks @bjcoombs! - Add modifyJSON function for safer file updates

• #1600 712a078 Thanks @esumerfd! - Add --no-banner to suppress the startup banner.

task-master-ai@0.42.0

Minor Changes

• #1533 6c3a92c Thanks @bjcoombs! - Add --ready and --blocking filters to list command for identifying parallelizable tasks
  • Add --ready filter to show only tasks with satisfied dependencies (ready to work on)
  • Add --blocking filter to show only tasks that block other tasks
  • Combine --ready --blocking to find high-impact tasks (ready AND blocking others)
  • Add "Blocks" column to task table showing which tasks depend on each task
  • Blocks field included in JSON output for programmatic access
  • Add "Ready" column to tags command showing count of ready tasks per tag
  • Add --ready filter to tags command to show only tags with available work
  • Excludes deferred/blocked tasks from ready count (only actionable statuses)
  • Add --all-tags option to list ready tasks across all tags (use with --ready)
  • Tag column shown as first column when using --all-tags for easy scanning

Patch Changes

• #1569 4cfde1c Thanks @bjcoombs! - Improve concurrency safety by adopting modifyJson pattern in file-storage

  • Refactor saveTasks, createTag, deleteTag, renameTag to use modifyJson for atomic read-modify-write operations
  • This prevents lost updates when multiple processes concurrently modify tasks.json
  • Complements the cross-process file locking added in PR #1566

• #1566 3cc6174 Thanks @bjcoombs! - Fix race condition when multiple Claude Code windows write to tasks.json simultaneously

  • Add cross-process file locking to prevent concurrent write collisions
  • Implement atomic writes using temp file + rename pattern to prevent partial writes
  • Re-read file inside lock to get current state, preventing lost updates from stale snapshots
  • Add stale lock detection and automatic cleanup (10-second timeout)
  • Export withFileLock and withFileLockSync utilities for use by other modules

  This fix prevents data loss that could occur when multiple Task Master instances (e.g., multiple Claude Code windows) access the same tasks.json file concurrently.

• #1576 097c8ed Thanks @Crunchyman-ralph! - Improve loop command error handling and use dangerously-skip-permissions

  • Add proper spawn error handling (ENOENT, EACCES) with actionable messages
  • Return error info from checkSandboxAuth and runInteractiveAuth instead of silent failures
  • Use --dangerously-skip-permissions for unattended loop execution
  • Fix null exit code masking issue

• #1577 e762e4f Thanks @Crunchyman-ralph! - Make Docker sandbox mode opt-in for loop command

  • Add --sandbox flag to task-master loop (default: use plain claude -p)
  • Preserve progress.txt between runs (append instead of overwrite)
  • Display execution mode in loop startup output

• #1580 940ab58 Thanks @Crunchyman-ralph! - Update Codex CLI supported models to match current available models

  • Remove deprecated models: gpt-5, gpt-5-codex, gpt-5.1
  • Add gpt-5.2-codex as the current default model
  • Add gpt-5.1-codex-mini for faster, cheaper option
  • Keep gpt-5.1-codex-max and gpt-5.2

task-master-ai@0.41.0

Minor Changes

• #1571 c2d6c18 Thanks @Crunchyman-ralph! - Add loop command for automated task execution with Claude Code

  New Features:

  • task-master loop command that runs Claude Code in a Docker sandbox, executing one task per iteration based on the selected tag
  • Built-in presets for different workflows:
    • default - General task completion from the Task Master backlog
    • test-coverage - Find uncovered code and write meaningful tests
    • linting - Fix lint errors and type errors one by one
    • duplication - Find duplicated code and refactor into shared utilities
    • entropy - Find code smells and clean them up
  • Progress file tracking to maintain context across iterations (inside .taskmaster/loop-progress.txt)
    • Remember to delete this file between loops to not pollute the agent with bad context
  • Automatic completion detection via <loop-complete> and <loop-blocked> markers

Patch Changes

• #1556 1befc6a Thanks @TheLazyIndianTechie! - fix: tolerate AI SDK versions without jsonSchema export

  Fallback to sanitized Zod schema handling when jsonSchema is unavailable, and
  align structured-output tests and registration perf thresholds to reduce CI
  failures.

  Also enforce sequential, unique subtask ids when regenerating subtasks during
  scope adjustment.

• #1553 226678b Thanks @Crunchyman-ralph! - fix: Remove .default() from Zod schemas to satisfy OpenAI strict JSON schema validation

  This fixes an issue where codex-cli provider (using OpenAI API) would fail with "Missing 'dependencies'" error during task expansion. OpenAI's structured outputs require all properties to be in the 'required' array, but Zod's .default() makes fields optional. The fix removes .default() from schemas and applies defaults at the application level instead.

• #1543 9a6fa1b Thanks @triepod-ai! - feat: Add tool annotations for improved LLM tool understanding

  Added MCP tool annotations (readOnlyHint, destructiveHint, title) to all 12 tools to help LLMs better understand tool behavior and make safer decisions about tool execution.

task-master-ai@0.40.1

Patch Changes

• #1523 fc1a79f Thanks @Crunchyman-ralph! - Codex cli Validate reasoning effort against model capabilities

  • Add provider-level reasoning effort validation for OpenAI models
  • Automatically cap unsupported effort levels (e.g., 'xhigh' on gpt-5.1 and gpt-5 becomes 'high')

• #1549 98087ac Thanks @Crunchyman-ralph! - Improve CLI startup speed by 2x

• #1545 a0007a3 Thanks @Crunchyman-ralph! - Smarter project root detection with boundary markers

  • Prevents Task Master from incorrectly detecting .taskmaster folders in your home directory when working inside a different project
  • Now stops at project boundaries (.git, package.json, lock files) instead of searching all the way up to the filesystem root
  • Adds support for monorepo markers (lerna.json, nx.json, turbo.json) and additional lock files (bun.lockb, deno.lock)

• #1523 fc1a79f Thanks @Crunchyman-ralph! - Improve json schemas for ai-related commands making it more compatible with openai models

  • Fixes #1541 #1542

• #1542 b817d6f Thanks @mdimitrovg! - Fixed vertex-ai authentication when using service account and vertex location env variable.

task-master-ai@0.40.0

Minor Changes

• #1538 a2d5639 Thanks @Crunchyman-ralph! - Added Gemini 3 Flash Preview model support for Google and Gemini CLI providers

• #1535 4d1ed20 Thanks @Crunchyman-ralph! - Add auto-detection for IDE profiles in rules command

  • tm rules add now opens interactive setup with detected IDEs pre-selected
  • tm rules add -y auto-detects and installs rules without prompting
  • Detects 13 IDEs: Cursor, Claude Code, Windsurf, VS Code, Roo, Cline, Kiro, Zed, Kilo, Trae, Gemini, OpenCode, Codex

• #1526 38c2c08 Thanks @eyaltoledano! - Add watch mode and compact output to list command

  • Add -w/--watch flag to continuously monitor task changes with real-time updates
  • Add -c/--compact flag for minimal task output format
  • Add --no-header flag to hide the command header
  • Support file-based watching via fs.watch for local tasks.json
  • Support API-based watching via Supabase Realtime for authenticated users
  • Display last sync timestamp and source in watch mode

Patch Changes

• #1538 a2d5639 Thanks @Crunchyman-ralph! - Improved model search in task-master models --setup to match both display names and model IDs

task-master-ai@0.39.0

Minor Changes

• #1521 353e3bf Thanks @Crunchyman-ralph! - Enhanced task metadata display for remote/team mode tasks

  • Tasks now display rich implementation guidance in team mode including:
    • Relevant Files: Files to create, modify, or reference with descriptions
    • Codebase Patterns: Coding patterns and conventions to follow
    • Existing Infrastructure: Code and utilities to leverage
    • Scope Boundaries: What's in and out of scope for the task
    • Implementation Approach: Step-by-step guidance
    • Technical Constraints: Requirements and limitations
    • Acceptance Criteria: Definition of done checklist
    • Skills & Category: Task classification and required expertise
  • How to see the new task details:
    1. Create a brief on tryhamster.com
    2. Generate the plan of the brief
    3. View subtasks

• #1525 1c2228d Thanks @Crunchyman-ralph! - Add tool search tool for Claude Code MCP server and enable deferred MCP loading

  • Added new tool search tool capabilities for the Taskmaster MCP in Claude Code
  • Running task-master rules add claude now automatically configures your shell (~/.zshrc, ~/.bashrc, or PowerShell profile) with ENABLE_EXPERIMENTAL_MCP_CLI=true to enable deferred MCP loading
  • Context savings: Deferred loading saves ~16% of Claude Code's 200k context window (~33k tokens for Task Master alone). Savings apply to all MCP servers, so total savings may be higher depending on your setup.

Patch Changes

• #1310 4b6570e Thanks @Crunchyman-ralph! - Fix Azure OpenAI provider to use correct deployment-based URL format
  • Add Azure OpenAI documentation page https://docs.task-master.dev/providers/azure

task-master-ai@0.38.0

Minor Changes

• #1461 9ee63e0 Thanks @Crunchyman-ralph! - Add operating mode filtering for slash commands and rules

  Solo mode and team mode now have distinct sets of commands and rules:

  • Solo mode: Local file-based storage commands (parse-prd, add-task, expand, etc.) plus common commands
  • Team mode: Team-specific commands (goham) plus common commands (show-task, list-tasks, help, etc.)

  Both modes share common commands for viewing and navigating tasks. The difference is:

  • Solo users get commands for local file management (PRD parsing, task expansion, dependencies)
  • Team users get Hamster cloud integration commands instead

  When switching modes (e.g., from solo to team), all existing TaskMaster commands and rules are automatically cleaned up before adding the new mode's files. This prevents orphaned commands/rules from previous modes.

  The operating mode is auto-detected from config or auth status, and can be overridden with --mode=solo|team flag on the rules command.

• #1461 9ee63e0 Thanks @Crunchyman-ralph! - Add Taskmaster slash commands for:

  • Roo
  • Cursor
  • Codex
  • Gemini
  • Opencode

  Add them with task-master rules add <provider>

• #1508 69ac463 Thanks @ben-vargas! - Added support for new OpenAI models with reasoning effort configuration:

  • gpt-5.1 (codex-cli & openai): supports none, low, medium, high reasoning
  • gpt-5.1-codex-max (codex-cli & openai): supports none, low, medium, high, xhigh reasoning
  • gpt-5.2 (codex-cli & openai): supports none, low, medium, high, xhigh reasoning
  • gpt-5.2-pro (openai only): supports medium, high, xhigh reasoning

  Updated ai-sdk-provider-codex-cli dependency to ^0.7.0.

task-master-ai@0.37.2

Patch Changes

• #1492 071dfc6 Thanks @Crunchyman-ralph! - Fix login issues for users whose browsers or firewalls were blocking CLI authentication

• #1491 0e908be Thanks @Crunchyman-ralph! - Add tm/ prefix to autopilot branch names

  • Team mode branches now follow the tm/<org-slug>/task-<id> naming convention for better organization.
  • Solves issue some users were having regarding not being able to start workflow on master Taskmaster tag

task-master-ai@0.37.2-rc.0

Patch Changes

• #1492 071dfc6 Thanks @Crunchyman-ralph! - Fix login issues for users whose browsers or firewalls were blocking CLI authentication

• #1491 0e908be Thanks @Crunchyman-ralph! - Add tm/ prefix to autopilot branch names

  • Team mode branches now follow the tm/<org-slug>/task-<id> naming convention for better organization.
  • Solves issue some users were having regarding not being able to start workflow on master Taskmaster tag

task-master-ai@0.37.1

Patch Changes

• #1477 b0199f1 Thanks @Crunchyman-ralph! - Improve auth-related error handling

• #1478 6ff330f Thanks @Crunchyman-ralph! - Increase page size of brief selection (interactive cli setup)