fix(rulestead): replace core-inherited changelog with the package's own

[szTheory]

What was wrong

packages/crosswake_rulestead/CHANGELOG.md was core's changelog, verbatim. It documented:

• the route policy DSL, bounded bridge contract, offline islands, commerce corridors
• SwiftPM / Maven Central shell-core publishing
• the RevenueCat provider seam
• the v9.0 brandbook

None of which are rulestead. Its compare links pointed at core's v0.1.0 / v0.1.2 tags rather than any crosswake_rulestead-v* tag, and it asserted "The current published Hex release is 0.1.2" — for a package that has never published at all.

Why it matters

CHANGELOG.md is in that package's files: allowlist:

files: ~w(lib mix.exs README.md LICENSE CHANGELOG.md)

So this file ships in the tarball and renders as rulestead's changelog on hex.pm. Hex publishes are irreversible. Had crosswake_rulestead gone out — and its Release PR #39 is open right now — core's entire release history would be permanently public under the companion's name.

This is the blocker I flagged on #84 as must-fix-before-any-future-publish. Fixing it now, while it's still cheap.

The fix

Replaced with a clean companion changelog mirroring the crosswake_rindle precedent: independent-versioning note + a single [Unreleased] extraction entry, plus a plain statement that the package is not yet on Hex.

Worth noting this was known: .planning/STATE.md records that rindle was deliberately given "a clean [Unreleased] skeleton, not core-inherited rulestead history" — rulestead's was simply never gone back and fixed.

Why it survived

No gate asserted on this file. grep across test/, script/, lib/, .github/ finds zero references to any per-package CHANGELOG.md content. Worth considering a publish-readiness check that a companion's changelog doesn't mention core-only surfaces — noted, not done here.

Tests

phase42_rulestead_companion + phase43_rulestead_advisory + phase130_extraction_guards + phase52_operator_truth → 22 tests, 0 failures.

🤖 Generated with Claude Code