PMO Platform para Claude Code
cody-hutson/pmo-platform
Gestão PMO e release pipeline modular para Claude Code com 13 estágios
Ver no GitHubPlataforma modular de gestão de projetos e pipeline de release integrada ao Claude Code. Oferece módulos independentes para PMOs (operações diárias, RAID logs, comunicação) e engenheiros (release management, pipeline de 13 estágios, deploys automatizados) com skills compartilhadas de governança.
Instalação
npx convez-ai cody-hutson/pmo-platform1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automação de pipeline de release com 13 estágios de intake a close
- Gestão de projetos e program management (RAID logs, status updates, comunicação)
- Versionamento de releases e bundle de backlog items com release notes automatizadas
Veredito

Solução robusta e bem documentada para equipes que precisam automatizar gestão PMO e release pipelines dentro do Claude Code. Requer investimento inicial em setup mas oferece composição modular real e integração profunda com workflows de DevOps.
README
A tradução para o português deste README ainda está na fila.
pmo-platform
A PMO platform with composable modules — operations, release, and core — designed to run with Claude Code (the CLI or the Desktop app).
What is it
pmo-platform is a modular-monolith codebase for program-management and release-pipeline work. Three modules cover distinct capability surfaces and serve two consumer audiences:
- PMO practitioners (Senior Program Managers, Technical Program Managers, PMO Directors) use the operations module for day-to-day program work — daily status updates, RAID Log maintenance, stakeholder communications, project artifact generation, transcript routing, and tracker upkeep.
- Platform builders (engineers, release managers, DevOps) use the release module for software-delivery management — bundling backlog items into versioned releases, driving the 13-stage pipeline from intake through close, executing deploys, and authoring release notes.
A shared core module hosts the security hooks, decision/discovery/review disciplines, schemas, deploy infrastructure, and the shared skills (prompt-builder, eval-writer, pmo-qa-auditor) consumed by both consumer modules.
Each module is independently consumable. A PMO practitioner can install the platform and use only the operations module; a platform-builder team can drive releases against any tracked repository using only the release module. Composing the two modules is also supported — the release module can ship releases for the operations module (and for itself).
Quick install
Requirements: Claude Code — the CLI or the Desktop app UI — plus macOS 12+, git, and python3. Three commands install pmo-platform on a fresh machine:
git clone https://github.com/cody-hutson/pmo-platform.git ~/Claude/pmo-platform
cd ~/Claude/pmo-platform
./install.sh
./install.sh bootstraps the workspace layout, resolves the operator-identity tokens, installs the core security hooks, lays down composition-surface seed files with marker fences (so future updates preserve any custom additions), and deploys all platform skills to the runtime path Claude Code expects. It is idempotent — running it again on an existing workspace updates rather than overwrites.
For prerequisites, the full step-by-step procedure, and troubleshooting, see docs/INSTALL.md. For the version-update procedure that preserves operator additions, see docs/UPDATE.md.
Getting started
After install, take the 5-minute taste: the first-task walkthrough at docs/GETTING_STARTED.md exercises one real skill end-to-end — invoke it, read the output, change your workspace identity, and re-invoke to watch context flow through.
Then go further with docs/FIRST_STEPS.md, which takes you from "I tried one skill" to running real work:
- Explore — the fastest way to learn the platform: open Claude Code in the repo and play Q&A with it ("explain the 13-stage pipeline", "what's a hub vs a spoke?"). The repo documents itself.
- Orient — the one mental model everything runs on: a typed work item → Triage → a versioned Milestone → a hub session that spawns a spoke per pipeline stage → close.
- Pick your track:
- PMO practitioner → configure your first project (
/project-initiator,PROJECT.md,delivery_approach) and use the daily operations skills. - Platform builder → hook up your GitHub repo (issues, a Milestone, the project board), then drive a release with the release module's hub.
- PMO practitioner → configure your first project (
FIRST_STEPS.md also covers the conventions worth knowing up front — governed-vs-ad-hoc changes, "recommend-then-you-decide" autonomy, evidence labels, and how to read a Decision Briefing.
After that, the module READMEs cover each module's full skill roster.
Module overview
| Module | Role | More |
|---|---|---|
operations/ | PMO operations capability — 13 skills for PMO practitioners | Module README |
release/ | Release pipeline / SDLC management — 6 skills for platform builders | Module README |
core/ | Shared kernel — hooks, schemas, shared skills, disciplines, deploy infrastructure | Module README |
docs/ | User-facing documentation — install, getting started, workspace setup, module APIs | Docs index |
For the consolidated cross-module API reference, see docs/module-apis.md.
Top-level layout:
pmo-platform/
├── operations/ # PMO operations module (13 skills, governance, templates)
├── release/ # Release pipeline module (6 skills, pipeline definitions, ADRs)
├── core/ # Shared kernel (hooks, schemas, disciplines, deploy infra)
├── docs/ # User-facing docs (INSTALL, GETTING_STARTED, workspace-setup, module-apis)
├── packages/ # .skill distribution artifacts per module
├── CHANGELOG.md # Keep a Changelog 1.1.0 format
├── LICENSE # BSL 1.1
└── README.md # This file
Architecture
The repository is laid out as a modular monolith: one repository, three modules, each with a documented Public API section in its README. Modules talk only through declared public surfaces; internal references stay inside their owning module. Extraction-readiness is preserved — if any module proves genuinely independent over time, it can be lifted into its own repository without rewriting consumer references.
The API-contract discipline buys two properties at low cost. First, the contract is explicit: any cross-module reference shows up in the consumer's reference list as a via-public-api entry, so drift between modules is visible at deploy time. Second, the kernel stays the kernel: anything core/ exports is consumed by both operations and release, so changes to the shared surface get reviewed against both consumers.
Cross-cutting capability lives in the core module as separately-named disciplines (decision, discovery, review) — small reference documents that operationalize how the platform treats tier classification, premise interrogation, and post-artifact verification. They're consumed by skills in both operations and release; their structural separation from any one skill keeps them composable.
For the architectural deep-dive (workspace layout, token-resolved config templates, where the security hooks live, how the deploy script bootstraps a fresh clone, how each module's skill roster maps to consumer audiences), see docs/workspace-setup.md.
License
Business Source License 1.1 (BSL 1.1).
- Non-production use is granted: personal use, educational use, and internal business operations (including use by employees, contractors, and agents for your own internal program and project management) are permitted.
- Production-as-commercial-offering requires a separate commercial license. The platform may not be incorporated into hosted services, managed services, or commercial software products without separate arrangement.
- The license converts to Apache 2.0 on the Change Date (2030-05-27), four years after first publicly available distribution of the licensed version.
See LICENSE for the full terms and the licensor contact for commercial-licensing inquiries.
Security
Found a vulnerability? Report it privately per the Security Policy — please don't open a public issue. The repo runs full-history gitleaks in CI, and native GitHub secret scanning + push protection are enabled.
Contributing
pmo-platform is a single-operator project under BSL 1.1. External pull requests are not accepted and write access is limited to approved collaborators — see CONTRIBUTING.md. You're welcome to fork for your own non-production use within the license and to read the code to learn the platform.
Status
Active development — not yet production-ready for public consumption. The downloadable surfaces (install.sh, the four onboarding documents under docs/, the three module READMEs, and the consolidated API reference) are in place and exercised end-to-end. The platform is functionally adequate for personal use and internal-business use per the BSL 1.1 grant above. A broader public-release readiness validation is in flight; the timing of that validation is not committed.
Acknowledgments
Built to run on top of Anthropic Claude Code and follows the Anthropic Skills convention for skill packaging. The Business Source License 1.1 framework is provided by MariaDB Corporation Ab. The CHANGELOG.md format follows Keep a Changelog 1.1.0 with semantic versioning adapted for the platform's release-milestone numbering.