{"id":32579558,"url":"https://github.com/anombyte93/prd-taskmaster","last_synced_at":"2026-06-13T03:05:32.786Z","repository":{"id":321263903,"uuid":"1080617301","full_name":"anombyte93/prd-taskmaster","owner":"anombyte93","description":"Zero-config goal-to-tasks engine for Claude Code (the Atlas engine). Graded PRD validation, dependency-ordered task graph, evidence-gated execution.","archived":false,"fork":false,"pushed_at":"2026-06-13T01:47:44.000Z","size":1249,"stargazers_count":509,"open_issues_count":2,"forks_count":52,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-06-13T02:22:03.278Z","etag":null,"topics":["agents","ai-development","claude-code","claude-skills","mcp","prd","product-management","product-requirements","requirements-engineering","taskmaster"],"latest_commit_sha":null,"homepage":null,"language":"Python","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/anombyte93.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","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":"2025-10-21T16:08:42.000Z","updated_at":"2026-06-13T01:47:46.000Z","dependencies_parsed_at":"2025-10-28T19:17:15.332Z","dependency_job_id":"3d5f93f0-708f-491c-95a7-64003503a7f6","html_url":"https://github.com/anombyte93/prd-taskmaster","commit_stats":null,"previous_names":["anombyte93/prd-taskmaster"],"tags_count":6,"template":false,"template_full_name":null,"purl":"pkg:github/anombyte93/prd-taskmaster","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/anombyte93%2Fprd-taskmaster","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/anombyte93%2Fprd-taskmaster/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/anombyte93%2Fprd-taskmaster/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/anombyte93%2Fprd-taskmaster/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/anombyte93","download_url":"https://codeload.github.com/anombyte93/prd-taskmaster/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/anombyte93%2Fprd-taskmaster/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34270418,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-13T02:00:06.617Z","response_time":62,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"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":["agents","ai-development","claude-code","claude-skills","mcp","prd","product-management","product-requirements","requirements-engineering","taskmaster"],"created_at":"2025-10-29T15:27:57.064Z","updated_at":"2026-06-13T03:05:32.781Z","avatar_url":"https://github.com/anombyte93.png","language":"Python","funding_links":[],"categories":["HarmonyOS","6 つのトラック"],"sub_categories":["Windows Manager","Tier 5：チーム / ロール別 Playbooks"],"readme":"# Turn any goal into shipped code.\n\n**prd-taskmaster** by [Atlas AI](https://atlas-ai.au) is an open-source engine for Claude Code\nthat takes a one-line goal, interviews you like a senior PM, writes a **graded,\nplaceholder-proof PRD**, compiles it into a **dependency-ordered task graph**, and executes every\ntask with **verification evidence** — so \"done\" means proven, not claimed.\n\nFree and MIT, forever.\n\nAtlas has four structural moats:\n\n- **cross-vendor fleet** — Claude, Codex, and Gemini run as separate quota pools instead of one\n  brittle model lane.\n- **Engine-enforced unfakable gates** — `validate-tasks`, evidence checks, and `SHIP_CHECK_OK`\n  make completion a deterministic state, not a claim.\n- **persistent vendor-neutral `tasks.json`** — your PRD, task graph, and execution state stay as\n  plain repo files that survive vendor swaps.\n- **token-economy cost ledger** — every orchestrated model call records routing, exit, latency,\n  and escalation so cheap models do cheap work and expensive models justify themselves.\n\n**Atlas speaks TaskMaster natively — but doesn't need it.** Existing TaskMaster projects get a\nmigration funnel: install `task-master-ai` only when you want the TaskMaster backend, while the\nnative backend keeps the same validated task graph available without that prerequisite.\n\n```\nGrade: GOOD  ▰▰▰▰▰▰▰▰▱▱  49/57 (86%) · 0 placeholders · 14 tasks parsed\n```\n\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)\n[![GitHub stars](https://img.shields.io/github/stars/anombyte93/prd-taskmaster)](https://github.com/anombyte93/prd-taskmaster/stargazers)\n[![works with](https://img.shields.io/badge/works_with-Claude%20%C2%B7%20Codex%20%C2%B7%20Gemini-8A2BE2)](https://atlas-ai.au)\n[![free engine](https://img.shields.io/badge/engine-MIT%20free%20forever-brightgreen)](#whats-open-whats-not)\n\n---\n\n## How it works\n\n```\ngoal → discovery interview → graded PRD → dependency-ordered task graph → verified execution\n```\n\n1. **Preflight** — detects your environment (native backend, optional TaskMaster backend, model CLIs, research) and configures it. Zero setup questions.\n2. **Discovery** — an adaptive, one-question-at-a-time interview captures your real constraints.\n3. **Generate** — writes a PRD, scores it against deterministic quality checks (letter grade), then parses it into a task graph with complexity scores and full subtask coverage.\n4. **Handoff** — detects what you have installed and recommends one execution mode.\n5. **Execute** — a CDD-gated loop implements each task and proves it with evidence, ending in a deterministic `SHIP_CHECK_OK` token.\n\n---\n\n## Quickstart\n\n90 seconds to your first run.\n\n### Path 1 — one-liner (recommended)\n\n```bash\ncurl -fsSL https://atlas-ai.au/install | bash\n# installs the skill + prd_taskmaster package\n# TaskMaster install is optional — unlocks the TaskMaster backend\n```\n\n### Path 2 — Claude Code plugin\n\n```bash\n# add the marketplace, then install the plugin\n/plugin marketplace add anombyte93/prd-taskmaster\n/plugin install prd\n\n# optional — unlocks the TaskMaster backend\nnpm install -g task-master-ai\n```\n\n### First run\n\nOpen any project in Claude Code and type:\n\n```\n/atlas      (or /prd:go, or just say: \"I want to build …\")\n```\n\nRequires Python 3.11+ and Linux / macOS / WSL. The free engine needs **no paid API key** — it\nuses the model CLIs you already have; an optional local research proxy can be plugged in\n(bring your own — not bundled). npm installs run a `postinstall` step that pip-installs the\nMCP server's Python deps (non-fatal warning if pip is unavailable).\n\n---\n\n## What \"verified\" means\n\nMost AI coding tools tell you a task is done. This one makes \"done\" provable:\n\n- **Graded PRDs.** Every spec is scored against deterministic checks (EXCELLENT / GOOD / ACCEPTABLE / NEEDS WORK). Placeholders (`TBD`, `{{...}}`, `TODO` — bare or bracketed) are a hard fail: the grade floors to NEEDS WORK and `validate-prd` exits non-zero.\n- **A real task graph.** Requirements become backend-neutral `tasks.json` tasks with dependencies, complexity scores, and full subtask coverage — not a flat checklist.\n- **Evidence-gated execution.** Each task is implemented and must produce execution evidence before it counts as done.\n- **A completion token you can trust.** `SHIP_CHECK_OK` is emitted only when every gate passes — and a single non-zero `Exit status` in any evidence file blocks it. It is structurally hard to fake. (One escape hatch exists for incident recovery: an explicit admin override flag that is audit-logged and marks the token `[OVERRIDE]` on stdout — never silent.)\n\n```\n┌─ atlas ── PHASE 3/4: GENERATE ─────────────────────────────┐\n│  Grade: GOOD  ▰▰▰▰▰▰▰▰▱▱  49/57 (86%)                      │\n│   ✓ 11 checks passed   structure · testability · metrics   │\n│   ⚠ 2 warnings   (quoted + located, not just counted)      │\n│   ✓ 0 placeholders     (TBD/TODO/{{...}} scan clean)       │\n│  Tasks: 14 parsed · 52 subtasks · dependencies mapped      │\n└────────────────────────────────────────────────────────────┘\n```\n\n---\n\n## Built for the token-shortage era\n\nEvery job runs on the cheapest model that can do it — and escalates only when a validator says\nit failed. One setting controls how aggressive that is:\n\n```json\n// .atlas-ai/fleet.json\n{ \"token_economy\": \"conservative\" }   // or \"balanced\" (default) / \"performance\"\n```\n\nTask decomposition and research run through the selected backend. Native mode works without a\nTaskMaster install; installing `task-master-ai \u003e= 0.43.0` unlocks TaskMaster's model-agnostic AI\n(any API you configure — Anthropic, OpenAI, Perplexity, Gemini, openai-compatible…) and isolated\nworkdir expansion when that backend is selected. Complexity 2 scaffolding gets a haiku-class\nmodel; the hardest long-running work gets the frontier model; nothing defaults to expensive.\nLocal telemetry (`economy-report`) shows your real success-rate and latency per model so the\nrouting gets smarter on YOUR workload — priors and sources in\n[`docs/product/MODEL-ECONOMY.md`](docs/product/MODEL-ECONOMY.md).\n\n## Free vs Atlas Pro\n\n| | Free (MIT) | Atlas Pro — $29/mo |\n|---|:---:|:---:|\n| Discovery interview (adaptive, one question at a time) | ✓ | ✓ |\n| Graded PRD validation + placeholder scan | ✓ | ✓ |\n| Dependency-ordered task graph (`tasks.json`) | ✓ | ✓ |\n| Verified solo execution — evidence required per task | ✓ | ✓ |\n| Model-agnostic: Claude / Codex / Gemini | ✓ | ✓ |\n| Parallel research fan-out | ✓ | ✓ |\n| **Token economy** — start cheap, escalate only on failure (conservative/balanced/performance) | ✓ | ✓ |\n| Optional **TaskMaster backend** expansion — any configured API, isolated workdirs | ✓ | ✓ |\n| Local cost telemetry + `economy-report` | ✓ | ✓ |\n| Adaptive routing auto-tuning from telemetry | — | ✓ (roadmap) |\n| **Atlas Fleet** — parallel waves of isolated workers, checker-gated merges, one final PR | — | ✓ |\n| Browser-verification MCP (UI proof, not just logs) | — | ✓ |\n| Secrets-vault MCP (keys never in your repo or prompts) | — | ✓ |\n| License \u0026 priority support | community | ✓ |\n\nThe engine is the whole engine — the free tier is **not a demo**. Pro adds the fleet orchestrator\nand bundled MCPs. [Get Atlas Pro →](https://atlas-ai.au/pro)\n\n---\n\n## Atlas Fleet (Atlas Pro)\n\nWhen a build is big enough to want overnight parallelism, **Atlas Fleet** lifts the same engine\nfrom one session to many. It splits your task graph into dependency **waves** of file-disjoint\nchunks, spawns model-agnostic workers (Claude / Codex / Gemini) in isolated **git worktrees**,\ncollects results through a durable inbox rather than trusting an agent's word, and merges waves\nsequentially into an integration branch behind a checker gate — never touching `main` until one\nfinal green PR.\n\nIt runs entirely on your machine; only the license key phones home. Cancel anytime — your specs\nand tasks are plain files in your repo, nothing is held hostage.\n\n[See Atlas Pro →](https://atlas-ai.au/pro)\n\n---\n\n## What's open, what's not\n\n**Open (MIT, free forever):** discovery, PRD validation, the task graph, and solo verified\nexecution — the entire engine in this repo. Read every line.\n\n**Closed (Atlas Pro):** the `atlas-launcher` fleet orchestrator and the two bundled MCPs\n(browser verification, secrets vault). The engine repo will never require a license key.\n\n---\n\n## FAQ\n\n**Will the free engine stay free?** Yes — MIT, and this repo will never require a key.\n\n**What happens if I cancel Pro?** Fleet locks again; everything free keeps working. Your specs\nand tasks are plain files in your repo.\n\n**Do I need a paid API key?** No. The engine uses the model CLIs you already have (Claude Code,\nCodex, Gemini); an optional local research proxy can be plugged in (bring your own — not bundled).\n\n**Do I need TaskMaster?** No. Atlas speaks TaskMaster natively but doesn't require it —\nNative Mode produces the same validated task graph (validate-tasks + enrich-tasks). Installing\n`task-master-ai \u003e= 0.43.0` unlocks the TaskMaster backend: Mode B auto-execute and\nnative-parallel expansion through TaskMaster's 13-provider model layer.\n\n**Which platforms?** Linux, macOS, and WSL. (Native Windows is not supported — the atomic state\nmachine uses POSIX file locking.)\n\n---\n\n## Contributing \u0026 License\n\nIssues and PRs welcome — see [CONTRIBUTING.md](CONTRIBUTING.md) and\n[CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md). Product and UX specifications live in\n[`docs/product/`](docs/product/).\n\nThe engine is **MIT licensed and will always be** — see [LICENSE](LICENSE). Upgrading from v3?\nSee [CHANGELOG.md](CHANGELOG.md). v3 remains available via `git checkout v3.0.0`.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fanombyte93%2Fprd-taskmaster","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fanombyte93%2Fprd-taskmaster","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fanombyte93%2Fprd-taskmaster/lists"}