Where this approach loses

Published by the sgit project, whose own skills are the eight scored on this site, and which sells the vault layer any registry described here would be built on. That is a conflict of interest worth naming plainly rather than leaving implicit.

What that means concretely

• The scorecard on the thesis page and the gaps on the shipped page are self-reported. Nobody outside the project verified the word counts, the coverage numbers, or the divergence measurement independently — though the divergence figure is at least recomputed on every build from a survey file that is itself published.
• The economy page’s scoring position — “provenance, evals, and an auditable graph, rather than gameable popularity” — happens to favour exactly the kind of infrastructure this project builds. That does not make the argument wrong, but it is not a disinterested argument either.
• The identity and permissions design (the identity page) proposes a frontmatter convention this project would be well placed to popularise if it shipped. Read the proposal on its merits, not its author’s incentive to be right.

What is not softened because of this

The one-green-cell scorecard, the diverging duplicate, and the zero-permissions, zero-evals, no-registry state of the estate’s own skills are published in full on the shipped page, unsoftened, specifically because a self-interested publisher who only publishes good news is not worth reading. The test of this disclosure is whether the unflattering numbers stayed in; they did.

This page follows the same disclosure pattern as every other site in the sgit.ai network.