MUSUBI experimental
Maximalist open-source SDD: EARS-format requirements, a nine-article constitution, CLI-enforced requirements-to-test traceability, and delta specs, installed into seven coding agents from one npx command.
Reviewed last month · Jul 18, 2026
- Stars
- 72
- Contributors
- 1
- Open issues
- 7
- Age
- 9 months
- Last push
- 8 months ago
- Latest release
- v6.3.1 Jan 1, 2026
View weekly commit data
| Week starting | Commits |
|---|---|
| Aug 23, 2026 | 0 |
| Aug 16, 2026 | 0 |
| Aug 9, 2026 | 0 |
| Aug 2, 2026 | 0 |
| Jul 26, 2026 | 0 |
| Jul 19, 2026 | 0 |
| Jul 12, 2026 | 0 |
| Jul 5, 2026 | 0 |
| Jun 28, 2026 | 0 |
| Jun 21, 2026 | 0 |
| Jun 14, 2026 | 0 |
| Jun 7, 2026 | 0 |
| May 31, 2026 | 0 |
| May 24, 2026 | 0 |
| May 17, 2026 | 0 |
| May 10, 2026 | 0 |
| May 3, 2026 | 0 |
| Apr 26, 2026 | 0 |
| Apr 19, 2026 | 0 |
| Apr 12, 2026 | 0 |
| Apr 5, 2026 | 0 |
| Mar 29, 2026 | 0 |
| Mar 22, 2026 | 0 |
| Mar 15, 2026 | 0 |
| Mar 8, 2026 | 0 |
| Mar 1, 2026 | 0 |
| Feb 22, 2026 | 0 |
| Feb 15, 2026 | 0 |
| Feb 8, 2026 | 0 |
| Feb 1, 2026 | 0 |
| Jan 25, 2026 | 0 |
| Jan 18, 2026 | 0 |
| Jan 11, 2026 | 0 |
| Jan 4, 2026 | 0 |
| Dec 28, 2025 | 26 |
| Dec 21, 2025 | 21 |
| Dec 14, 2025 | 2 |
| Dec 7, 2025 | 163 |
| Nov 30, 2025 | 33 |
| Nov 23, 2025 | 42 |
| Nov 16, 2025 | 33 |
| Nov 9, 2025 | 0 |
| Nov 2, 2025 | 0 |
| Oct 26, 2025 | 0 |
| Oct 19, 2025 | 0 |
| Oct 12, 2025 | 0 |
| Oct 5, 2025 | 0 |
| Sep 28, 2025 | 0 |
| Sep 21, 2025 | 0 |
| Sep 14, 2025 | 0 |
| Sep 7, 2025 | 0 |
| Aug 31, 2025 | 0 |
Metrics updated Aug 26, 2026 · collected automatically from the GitHub API
Core approach
Rigor turned all the way up. Work is governed by nine "immutable" constitutional articles — test-first, EARS requirement syntax with REQ-XXX-NNN IDs, 100% requirements → design → code → test traceability — enforced by a constitution-enforcer skill and audited by dedicated CLIs (musubi-trace, musubi-gaps). Spec artifacts live under storage/ (specs, design, tasks, validation reports), project memory under steering/, and every generated document is produced bilingually in English and Japanese. The README credits six predecessor frameworks, taking constitutional governance from Spec Kit and delta specs from OpenSpec, and orchestrates roughly 25 specialized agent skills across the lifecycle.
Workflow
npx musubi-sdd init— scaffold commands, steering, and templates for the chosen agents (ormusubi-onboardto reverse-engineer steering docs from an existing codebase)/sdd-steering— generate project memory (structure, tech stack, product context, rules)/sdd-requirements— EARS-format requirements with unique requirement IDs/sdd-design— design document with C4 model and ADRs/sdd-tasks— task breakdown/sdd-implement— test-first implementation/sdd-validate— constitutional-compliance check- Brownfield track —
/sdd-change-initwith automatic impact analysis, then/sdd-change-applyand/sdd-change-archive, usingADDED/MODIFIED/REMOVEDrequirement deltas - Per-stage review gates —
/sdd-review-requirements,/sdd-review-design,/sdd-review-implementation
Supported tools
- Claude Code
- GitHub Copilot
- Cursor
- Gemini CLI
- Codex
- Qwen Code
- Windsurf
Strengths
- The most formally rigorous open SDD tool tracked here — traceability is enforced by real tooling (
musubi-tracecoverage thresholds, gap detection, impact analysis), not just by prompts - Serious brownfield story: OpenSpec-style requirement deltas plus
musubi-onboard, which generates steering documents from an existing codebase before iteration begins - Seven coding agents supported from one installer, each in its native command format plus an
AGENTS.mddefinition for non-Claude platforms - Unusually extensive documentation for the project's size, with English and Japanese mirrors of every rule and guide
Limitations
- Marginal adoption and apparently stalled — a single maintainer, no commits or releases since January 2026, and the author has already started a successor project (MUSUBIX)
- Heavy ceremony: nine constitutional articles, EARS syntax, C4 diagrams, and a traceability matrix before code — and some articles (library-first, a CLI for every library, no mocks) are opinionated to the point of being unworkable for many codebases
- The full experience is Claude Code-only by the project's own platform comparison — auto-invoked skills, orchestration, and traceability auto-audit don't reach the other six agents, which get static prompt files
- Single-author velocity over polish: 75 npm versions across six major versions in the first 46 days, with internal inconsistencies in the docs (conflicting skill counts, license vs project plan)
Notes
MUSUBI is the maximalist pole of the SDD spectrum — 2026 framework surveys use it as the rigor-extreme datapoint opposite the lightweight, high-adoption kits. It is explicitly a synthesis play: the README credits six predecessors, taking constitutional governance from Spec Kit and delta specs from OpenSpec, then turning both up — the constitution ships in a file literally marked “Immutable”, and traceability is a CLI-enforced 100% requirement rather than an aspiration. As evidence for the observatory it cuts both ways: proof that full requirements → design → code → test traceability can be tooled end-to-end by a single author, and — with a few dozen stars, a repo quiet since January 2026, and that author already building a successor — a datapoint that maximal rigor has not yet found an adopter base to sustain it.