Acknowledge to @OpenAI Thanks for delivering this skills into reality
I've just shipped the most complete release of the Codex skills bundle yet, with a real demo, a restructured onboarding story, and 33 documented plugins and features across the full skill surface.
Repository: https://github.com/JustineDevs/meta-architect
Why this release matters v0.1.6 was about coherence getting the packaging, identity, and install story right. v0.1.13 is about showing the product clearly and making it easier to trust.
The gated lane sequence hasn't changed. But everything around it has: the manager is smarter, there are more skills to reach for, the install surface covers more platforms, and the README now leads with a live demo and a capability table instead of release-maintainer details.
If v0.1.6 was "ready enough for people who want to use it in anger", v0.1.13 is "ready enough for people who want to understand it in 30 seconds and then use it in anger".
What v0.1.13 focuses on A real demo, not static copy
The README now opens with a live demo GIF that shows install, $maestro orchestration, Obsidian brain context, learning reliability, environment awareness, Ralph handoff, and release verification — with real runtime functions, not screenshots of documentation.
Cleaner onboarding path
The README now leads with:
Install command
Demo media
Capability summary table
Grouped feature inventory (expandable)
Long-form release-maintainer details are collapsed into blocks. A first-time viewer can follow the product story without wading through packaging rules.
"What You Get"
capability table A new top-level table makes the full surface scannable at a glance:
An expandable section now lists every plugin and feature across six categories: Core & Orchestration, Memory & Knowledge, Intelligence & Learning, Code Quality & Testing, Security & Compliance, Architecture & Methodology, DevOps & Observability, and Extensibility.
How you use it
Visit Setup Guide
Who this helps For developers
The demo shows the full workflow end-to-end before you install anything
The helper skills give you targeted tools between gates — alignment, diagnosis, test scaffolding, cleanup — without disrupting release state
- ma setup / ma bootstrap / ma doctor remove the manual setup friction that tripped people up in earlier versions
For new architects
- $maestro as a bounded autonomous manager means you can ask it to choose the next step rather than tracking the sequence yourself
- The "What You Get" table and 33-plugin inventory give you a map of the surface before you start
For experienced architects
The gate model is explicit: CLEAR, APPROVED, VERIFIED, PARTIAL, GREEN,
RED, WAIVED, LOCKED, READY — each status has a defined meaning and a
defined effect on what is allowed next
Evidence must come from repository-form GitMCP endpoints. A generic docs endpoint does not count as VERIFIED for build unlocking
The Redaction Gateway, MCP policy, and exposure catalog give you explicit control over what context crosses trust boundaries
I'm especially interested in feedback from people who:
- are running the gated sequence on real projects and hitting edge cases
- want to use the helper skills ($align, $diagnose, $tdd, $cleanup) and have opinions on where they fall short
- are on Linux and trying the native package install path for the first time
- watched the demo and found something that didn't match what the runtime actually does
If that's you, the repo is here: https://github.com/JustineDevs/meta-architect