Decision: publish Date: 2026-09-26 By: MaverickHQ

The publication decision under R7, recorded by the author after reviewing the repository as it stands at this commit ("this is good, reviewed and ready to set to publish"). What is published is the repository and its pages as a static site: the lab, the record and the results, as README.md states under What is published here; the market data, the author's configuration and every credential stay out, by construction.

Before this record: the whole-tree publication scan reports zero unresolved hits, with 308 kept by the author's recorded decisions in tools/publication-decisions.toml (M15.3); make site renders every page and document through the publication check, clean (M15.2); the standing checks are green on the CI machine, the clean-machine job among them (M15.4). The history is published with the tree and is unchanged: every verdict stamps its commit.

What this record permits: the pages workflow, which reads the three lines above and deploys nothing without them, and the repository's visibility, which is the author's setting on GitHub and not this file's to change. The lab's product is its refusals; nothing published here places an order or recommends a number.

Amendment, 2026-10-03 (appended; nothing above is edited). The decision to publish stands. One sentence above is superseded: "The history is published with the tree and is unchanged."

The author decided that the public repository carries no history from before publication. Two early commits carried a private address in their metadata, and the author's words were "I don't want my email public". Of the ways to keep it private the author chose a snapshot — "a new branch with no history", published as the public main — over rewriting those commits. The public repository therefore begins at a single commit that holds the lab's tree as it stood, and it was created fresh, so that it has never contained the earlier commits.

What that changes, and what it does not. The full history stays whole in a private repository, archived and read-only. The commit ids the task list cites, and the ids stamped on verdicts as engine_sha, name commits in that private history; a reader of the public repository cannot inspect them. Nothing else moved: the three hash-chained Registers, the forty-six decisions and the dated log are published as they were, and the Registers' chains verify. From the snapshot on, history is never rewritten; a ruleset refuses a force-push or a deletion on main and a moved release tag, for anyone.

The pre-publication audit of 2026-09-27 listed three other things that publishing exposes, none of them a credential: the venue's name and a help-centre article number in dated evidence, and the donor's figures in a test fixture. The author confirmed each as kept, under the decisions already recorded in tools/publication-decisions.toml.

The repository was set public by the author on 2026-10-03, and the site was deployed the same day.