Authority rows: the release-prep bump follows the family norm

[johnnyt]

Rewords two rows of the authority table in CLAUDE.md and the sentences in .claude/wurk/release.md that followed the old bump row. Docs-only; no package code, no version, no tag.

The version-bump row

It said: a version bump on a release bead's branch is allowed on "an operator-authorized release bead, inside a campaign carrying the operator's explicit consent". It now says: "a release bead the operator has named (in the campaign plan or their own words); the bump is release prep, the family norm, and no campaign consent has to name it - its tag follows the tagging row once the prep is merged to origin/main, and the publish stays the operator's". This restates the file's own "Release preps" paragraph (the version bump and the tag of a release prep are the family norm; the tag once the prep is merged; mix hex.publish the operator's), ruled by the operator, 2026-09-27. It grants nothing that paragraph does not: the forbidden cell ("on any other bead, on main, or when the operator has not named this repo's release bead") is byte-identical, and the tagging row and the release row are untouched.

The tracker-push row

It said: bd dolt push is "the operator's call", and inside a campaign that spans mirrored trackers the conductor pushes those atomically. It now says: bead state changed locally and the git side of the same change has already reached origin; inside a campaign, the conductor pushes (atomically across the campaign's trackers); still unauthorized as a way to publish beads for work that is not on origin/main yet. Ruled by the operator, 2026-09-29, in these words; nothing beyond them is granted.

.claude/wurk/release.md

• The bullet quoting the version-bump row now quotes the new trigger; its "still unauthorized" quote is unchanged.
• The push-and-request paragraph dropped "and the same consent the version-bump row above needs": the push row still needs the campaign consent, the bump no longer does.
• "on a named release bead's branch, under a campaign consent that names it" now reads "on the branch of a release bead the operator has named".

The publish sentences are unchanged. .claude/wurk/commit.md's "Version bump: never" section already said "a release bead the operator has named" and is unchanged.

The provenance line

The merge exception's "(Recorded 2026-09-01 by the operator, ...)" line drops an internal reference that ran across a line break; the date and the attribution stay.

Review

In-turn review against the bead's acceptance. Every other table row is byte-identical: a diff of the lines starting with | in CLAUDE.md between origin/main and this branch shows exactly two changed lines, the two rows above. Each new sentence was read against what it restates (the "Release preps" paragraph for the bump row and the release.md sentences, the operator's wording for the tracker-push row); none widens a merge row, the tag row, the bd close row or the publish row. The public-tree id scan's joined pass reported one line on origin/main (the split reference above) and none on the edited files.

Gate

No path under the manifest's gate.build_paths is touched, so there is no gate to run (.claude/wurk/mr.md: a branch touching no gated paths has no gate to run). No changelog fragment: changelog.d/README.md excludes documentation and agent tooling changes.