Skip to content
/
Tier B A4 Markdown scaffold Variant

AI Task Master (James-Cherished-Inc)

ai-task-master-james · James-Cherished-Inc/AI-task-master · ★ 21 · last commit 2025-03-29

Fork of claude-task-master adapted for Roo Code, Cline, and OpenRouter multi-model API routing.

Best whenOpenRouter's 50+ free models and reliable rate limits make it a better default API choice than Anthropic direct for task management workloads.
Skip ifHard-coding model names, Skipping testStrategy when generating tasks
vs seeds
taskmaster-aiseed (Archetype 3 — file-based JSON task management with CLI). Fork-specific additions: OpenRouter as multi-model provid…
Primitive shape
No installable primitives
00

Summary

AI Task Master (James-Cherished-Inc) — Summary

AI Task Master by James Cherished Inc is a fork of the canonical claude-task-master (eyaltoledano) adapted for Roo Code, Cline, and OpenRouter. The repo has 4 branches targeting different AI tool combinations: cursor (tracks upstream), roo-as-custom-mode (Roo Code via custom mode + npm package), Roo-AI-Task-Master (modified npm package for Roo), and Roo+OpenRouter (OpenRouter as API provider replacement). The main branch README is a comprehensive guide explaining the branch strategy. Core functionality is inherited from eyaltoledano's taskmaster-ai: CLI-driven PRD parsing, dependency-aware task management (tasks.json), and AI-guided implementation. The James fork's key additions are: OpenRouter as a multi-model provider option (replacing Anthropic + Perplexity with 50+ free models), Roo Code custom mode integration, and Cline rule file (.clinerules-taskmaster). 21 stars, no license, last commit 2025-03-29 (dormant).

Differs from seeds: A derivative of taskmaster-ai seed (Archetype 3 — MCP-anchored toolserver). The canonical canonical: false should be set; this is not the upstream. Unlike the seed which targets Cursor primarily, this fork targets Roo Code and Cline with OpenRouter multi-model support. The OpenRouter integration enables genuine multi-model routing (SOTA models, cost-optimized fallback) that the original seed lacks.

01

Overview

AI Task Master (James-Cherished-Inc) — Overview

Origin

Forked from eyaltoledano/claude-task-master by @JamesCherished. Fork motivation: make taskmaster work with Roo Code and OpenRouter. Last active 2025-03-29.

"Initially developed for Cline with Anthropic's Claude by @eyaltoledano" "Forked by @JamesCherished to work with Roo Code and Cline."

Branch strategy

Branch Target Tool Key Difference
cursor Cursor Tracks upstream eyaltoledano
roo-as-custom-mode Roo Code Prompts in Task Master Prompts.md + npm package
Roo-AI-Task-Master (preferred) Roo Code Modified npm package for Roo-native experience
Roo+OpenRouter Roo Code OpenRouter replaces Anthropic/Perplexity API

OpenRouter value proposition

"Benefits: More reliable API access, Better rate limits, Consistent performance, Lower costs and better caching, 50+ free models, most SOTA models"

Roo Code integration

  • Architect mode is embedded: if no documentation exists at session start → automatic design phase
  • If documentation/plans exist → immediately resumes implementation
  • Custom mode via Task Master Prompts.md in .vscode/

Core workflow (inherited from upstream)

PRD → parse-prd → tasks.json → list → next → implement → set-status:done

Scope

21 GitHub stars on the main branch; this is a power-user fork, not a general audience distribution.

02

Architecture

AI Task Master (James-Cherished-Inc) — Architecture

Distribution

Fork of npm package task-master-ai (upstream: eyaltoledano). The Roo-AI-Task-Master branch ships a modified npm package to git clone + run locally.

Install

# Roo-AI-Task-Master branch (preferred)
git clone https://github.com/James-Cherished-Inc/AI-task-master -b Roo-AI-Task-Master
npm install

# Or use npm package
npm i -g task-master-ai
task-master init

Configuration (.env)

ANTHROPIC_API_KEY=...    # Option 1: Claude
PERPLEXITY_API_KEY=...   # Option 2: Perplexity
OPENROUTER_API_KEY=...   # Option 3: OpenRouter (fork-specific)
AI_PROVIDER=CLAUDE|PERPLEXITY|OPENROUTER
OPENROUTER_PARSING_MODEL=anthropic/claude-3-sonnet
OPENROUTER_RESEARCH_MODEL=anthropic/claude-3-sonnet
MODEL=claude-3-7-sonnet-20250219
MAX_TOKENS=4000
TEMPERATURE=0.7

Target AI tools

  • Roo Code (primary, custom mode)
  • Cline (.clinerules-taskmaster)
  • Cursor (cursor branch)
  • OpenRouter (multi-model provider)

Runtime requirements

  • node>=14
  • npm/yarn/pnpm
  • Anthropic API key OR OpenRouter API key
03

Components

AI Task Master (James-Cherished-Inc) — Components

CLI commands (inherited from upstream, ~20)

Command Purpose
task-master init Initialize project
task-master parse-prd <file> Parse PRD → tasks.json
task-master list List all tasks
task-master next Show next task (dependency-aware priority)
task-master show <id> Show task details
task-master expand --id=<id> Break task into subtasks
task-master expand --all Expand all pending tasks
task-master analyze-complexity AI complexity analysis (1–10 scale)
task-master complexity-report View complexity report
task-master set-status --id=<id> --status=<status> Update task status
task-master update --from=<id> --prompt=<text> Update future tasks for implementation drift
task-master generate Generate individual task files from tasks.json
task-master add-task --prompt=<text> Add task via AI
task-master add-dependency Add task dependency
task-master validate-dependencies Check dependency validity
task-master fix-dependencies Auto-fix invalid dependencies

Fork-specific commands

Command Purpose
npm run expand-research Research-backed task expansion (OpenRouter)
npm run analyze-complexity-research Research-backed complexity analysis (OpenRouter)
OPENROUTER_API_KEY=... npm run parse-prd-openrouter <file> OpenRouter PRD parsing

State files

  • tasks.json — central task registry with deps, priorities, details, subtasks, testStrategy
  • scripts/task-complexity-report.json — AI complexity analysis output
  • tasks/task-<id>.md — individual task files (generated)

AI integration files

  • Task Master Prompts.md — Roo Code custom mode prompts
  • .clinerules-taskmaster — Cline rules file
05

Prompts

AI Task Master (James-Cherished-Inc) — Prompts

Prompt 1: Task structure definition (from README)

Technique: JSON schema definition as execution contract

{
  "id": 1,
  "title": "Initialize Repo",
  "description": "Create a new repository, set up initial structure.",
  "status": "pending",
  "dependencies": [1, 2],
  "priority": "high",
  "details": "Use GitHub client ID/secret, handle callback, set session token.",
  "testStrategy": "Deploy and call endpoint to confirm 'Hello World' response.",
  "subtasks": [{"id": 1, "title": "Configure OAuth", ...}]
}

Key design: testStrategy is a per-task field that defines verification criteria at task creation time — not at implementation time. This is prompt-chaining: the AI that generates the task also specifies how to verify it.

Prompt 2: Prompt-chaining pattern (from README, analyze-complexity → expand)

Technique: Output of one AI call becomes prompt for the next AI call

From README:

"The generated [complexity] report contains: Complexity analysis for each task (scored 1–10), Recommended number of subtasks based on complexity, AI-generated expansion prompts customized for each task, Ready-to-run expansion commands directly within each task analysis."

# Step 1: Generate complexity report with expansion prompts
task-master analyze-complexity --research

# Step 2: expand consumes the expansionPrompt written by step 1
task-master expand --id=8
# Uses recommended subtask count and AI-generated expansion prompt

This is the same prompt-chaining pattern identified in the upstream taskmaster-ai seed — preserved and extended with research-backed variants in this fork.

09

Uniqueness

AI Task Master (James-Cherished-Inc) — Uniqueness

Differs from seeds

A derivative of the taskmaster-ai seed (Archetype 3 — MCP-anchored toolserver, file-based task management). The James fork's primary additions: (1) OpenRouter as an API provider alternative, enabling genuine multi-model routing with 50+ models including free tiers — the upstream seed's multi-model support is more limited; (2) Roo Code custom mode integration via Task Master Prompts.md; (3) Cline support via .clinerules-taskmaster. Unlike the upstream which targets Cursor primarily, this fork is explicitly Roo/Cline-first. The Roo integration's architect-first auto-detection (check for docs → enter architect mode if none, resume coding if docs exist) is not in the upstream.

Positioning

A power-user fork for developers who prefer Roo Code or Cline over Cursor, or who want OpenRouter's broader model ecosystem. Not a general distribution — assumes familiarity with the upstream taskmaster-ai.

Observable failure modes

  1. Dormant since 2025-03-29: Upstream has evolved significantly; this fork may be outdated.
  2. No license: Cannot be used in commercial contexts without license clarification.
  3. Branch complexity: 4 branches with different installation paths creates confusion for new users.
  4. Roo Code platform risk: If Roo Code changes its custom mode API, the integration breaks.
  5. Fork drift: The cursor branch "aims to be up-to-date" but may lag; the Roo branches have diverged further.
04

Workflow

AI Task Master (James-Cherished-Inc) — Workflow

Main workflow (PRD → implementation)

Phase Command Artifact Gate
Parse requirements task-master parse-prd prd.txt tasks.json Human review of generated tasks
Review tasks task-master list Task overview —
Analyze complexity task-master analyze-complexity [--research] scripts/task-complexity-report.json —
Expand complex tasks task-master expand --all Subtasks in tasks.json —
Implement task-master next → implement → test Code Per-task testStrategy
Update status task-master set-status --id=<id> --status=done tasks.json updated —
Handle drift task-master update --from=<id> --prompt="Changed approach" Future tasks updated —

Prompt-chaining pattern

analyze-complexity writes expansionPrompt per task → expand --id consumes it. One stage's output is the next stage's prompt.

Roo Code workflow

  1. Start Roo task with Task-Master custom mode
  2. If no docs: Roo enters Architect mode → creates docs
  3. If docs exist: Roo runs task-master list, resumes pending tasks
  4. Each task cycle: task-master next → implement → task-master set-status

Approval gates

  1. PRD review before parsing
  2. Human review of tasks.json after parse-prd
  3. Per-task verification against testStrategy before set-status:done
06

Memory Context

AI Task Master (James-Cherished-Inc) — Memory & Context

State files

  • tasks.json — central task registry (file-based, project-scoped)
  • scripts/task-complexity-report.json — AI complexity analysis
  • tasks/task-<id>.md — per-task detail files

Persistence

File-based, project-scoped. All state survives session restarts via tasks.json. task-master list always shows current state.

Cross-session handoff

Yes — task-master list at session start always shows resumable state. The Roo Code integration explicitly checks for existing tasks at session start to determine architect vs code mode.

Implementation drift handling

task-master update --from=<id> --prompt="..." rewrites future tasks when implementation diverges from plan — forward-looking memory update without destroying completed work.

07

Orchestration

AI Task Master (James-Cherished-Inc) — Orchestration

Multi-agent

No — single agent executes all tasks sequentially.

Orchestration pattern

sequential — tasks executed in dependency order, one at a time.

Multi-model

Yes (fork addition) — OpenRouter provider enables multi-model routing:

Role Model
PRD parsing OPENROUTER_PARSING_MODEL (default: anthropic/claude-3-sonnet)
Research OPENROUTER_RESEARCH_MODEL (default: anthropic/claude-3-sonnet)
General MODEL env var

Prompt-chaining

Yes — analyze-complexity output feeds expand input (same as upstream seed).

Isolation

None — operations in-place on local filesystem.

Execution mode

interactive-loop — Roo Code custom mode continuously checks task state and implements.

Supports BYOK

Yes — Anthropic, Perplexity, or OpenRouter API keys.

08

Ui Cli Surface

AI Task Master (James-Cherished-Inc) — UI & CLI Surface

Dedicated CLI binary

Yes — task-master CLI (inherited from upstream)

  • Package: task-master-ai
  • Is thin wrapper: No — the CLI does the actual PRD parsing and task management via AI calls
  • Key subcommands: init, parse-prd, list, next, show, expand, analyze-complexity, complexity-report, set-status, update, generate, add-task, add-dependency, validate-dependencies, fix-dependencies

Local web dashboard

None.

IDE integration

  • Roo Code (custom mode via Task Master Prompts.md + .vscode/)
  • Cline (.clinerules-taskmaster)
  • Cursor (cursor branch)

Observability

  • tasks.json — human-readable task state
  • tasks/task-<id>.md — per-task detail files
  • scripts/task-complexity-report.json — complexity analysis output

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k
A4 Markdown scaffold

Provides a full agile delivery lifecycle with named expert-persona AI collaborators that elicit the human's best thinking rather…

Agent OS ★ 4.6k
A4 Markdown scaffold

Extracts implicit codebase conventions into token-efficient markdown standards files and injects them selectively into AI agent…

Claude Conductor ★ 367
A4 Markdown scaffold

Gives Claude Code a persistent, cross-linked, auto-analyzed documentation system so it retains codebase context across sessions.

Spec-Driver (Greenfield Spec-Driven Development) ★ 25
A4 Markdown scaffold

Prevents spec rot in AI-assisted development by making implementation changes flow back into evergreen, authoritative specs via…

Anthropic Knowledge Work Plugins ★ 16k
A4 Markdown scaffold

Role-specialized plugin bundles with live MCP connectors that turn Claude into a domain expert for enterprise knowledge workers.

Codex Integration for Claude Code (skill-codex) ★ 1.3k
A4 Markdown scaffold

Single Claude Code skill that handles Codex CLI invocation correctly (stdin blocking, thinking token suppression, session resume)…