chore(submodules): promote PMOVES-crush -> db4e2ef05 (catwalk TensorZero activation)

[POWERFULMOVES]

What

Promotes PMOVES-crush 76521dc6 → db4e2ef05 after fork PR #15 activated the pmoves-catwalk submodule build: replace charm.land/catwalk => ./pmoves-catwalk in go.mod + submodule-init in every go-toolchain workflow (lint via meta's pre-run hook). Published fork binaries now carry the TensorZero provider — closing the 2026-07-18 trail deferral (Left Behind #1).

• Compare ahead (clean fast-forward); fork PR merged with a merge commit
• CI-proven on the fork: build 3-OS + lint + security all success WITH the replace active (the exact configuration that was previously impossible without the submodule lane)
• Locally verified on go1.26.6: build exit 0, tidy idempotent, schema command valid
• Register CLAIM+RELEASE ride along, incl. the disclosed ~50min fork Actions event-pipeline stall (recovered, documented on the fork PR)

💘 Generated with Crush

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: c805a1c0-3df1-444d-837d-cbb8cdf1382c

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0bca679dc2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[POWERFULMOVES]

P1 addressed: fork PR #16 merged (bab9f13e), and this PR's gitlink advanced to that tip. You were right that the NOTE-only mitigation left a latent failure. What landed: nightly.yml inlined as a 3-OS go build+test with submodule init (the new file verified by a successful workflow_dispatch run on the branch — signal survives with zero secrets, and it exercises the replace); release.yml became an explicit not-configured guard that fails with instructions instead of a confusing goreleaser error. Evidence base recorded in the PR: goreleaser-pro has never succeeded on this fork (the single 2026-08-05 snapshot run failed on the missing GORELEASER_KEY; nightly had been startup-failing on the same secret for weeks) — so no working path was removed, and the operator decision (publish fork binaries or not) is now visible instead of buried in a startup_failure.