Open source

How this is built, and what is public

Most of what is interesting here is not the licence but the method: this is developed against a real vault, with real meetings in it, and published anyway. That takes a mechanism rather than good intentions.

What is public

The skills, the contract, the change-request index and the release history: tandregbg/core-claude-skills. Contract 41, core-skills 1.89.4, 11 invocable skills.

What is not, and why that is stated rather than hidden

Five of the nine components are local tools in private repositories. They are described in this manual anyway — with each page saying plainly that the part is not published — because a manual that only covers what a reader can install is a manual that misrepresents the system.

There is one more thing worth knowing: a snapshot of the dashboard was published in March and has not been touched since. It is public, it is old, and the current version is not that snapshot.

Developed on production data, which is the interesting part

The repository is public. The vault it operates on is real: real people, real organisations, real infrastructure. Development runs on that live data by design — the skills improve by observing real usage — which makes the boundary between the two non-negotiable rather than a matter of care.

The membrane is one-way and narrow. Experience from the vault reaches the repository only through the change-request pipeline, rewritten generically; nothing else crosses. A guard runs before every push and scans every added line.

The part worth borrowing is the inversion that makes it watchable: real names cannot all be enumerated, but invented ones can. Every name-like token in the repository must match an allow-list of names that are known to be fictional. An unknown token blocks the push. Adding a line to that list is a conscious act — confirm the name is invented, never borrowed.

The licence

From ecosystem.yaml, contract 41 · core-skills 1.89.4 · read at build 2026-10-08