docs(register): close one lane honestly, acknowledge five delivered-but-unreleased

[POWERFULMOVES]

Six expired-and-unreleased claims were each blocking a future claimant while nobody had said they were still on it. Five were already delivered and simply never released. One was mine and was genuinely unfinished.

Mine — released with the real verdict

docs/reclaim-cipher-token-model: still needed, never started, now split. Claimed twice (#2824, #2864) and both PRs were register-only — #2864 was 6+ 0-, one file. Any audit keying on PR state reads two merged PRs as closure.

It kept expiring because one claim held two lanes and the expensive half gated the cheap half. Cheap half landed as #2988. The expensive half — unhashed storage on a live auth table, auth.ts:62-64 in a submodule, bootstrap's six-scope shared identity (the active mode on Z890) — is named and left unclaimed, because each needs operator sequencing and bundling them is what killed the lane twice.

Five backfill acknowledgements — not owner-signed closures

Lane	Owner	Evidence
fix/register-read-path	B850	#2894 eb03f993c
fix/cipher-mcp-transport-build-pin	B850	#2875 87863ce84
fix/embedding-path-restore	B850	#2895 accd3e406
feat/fleet-sentinel-console	HERMES	#2934 a236c2ff8
fix/lite-bootstrap-no-cuda	SPARK	#2961 4a97fb948

Each carries filed-by: Z890-CLAUDE, the merge SHA and timestamp, and states that the owner-authored RELEASE and signed ACK remain owed. Attribution stays with the owner; only the signature is mine.

That split is the whole point. AGENT_IDENTITY_PROPOSAL_2026-09-04.md §4 names "written about and never writes" as the failure mode — an agent denied the means to declare identity and then judged for not declaring it. A backfill must not quietly become another agent's ledger entry authored by me.

Citation correction

The precedent for this (#2960, the Mavis row) cites "AGENTS.md:L60-62, co-owner has no authority to close another owner's claim." That clause is not in AGENTS.md, and git log -S over AGENTS.md history returns empty — it has never been there.

The practice is sound and is followed here; the phantom citation is not propagated. Grounded instead on the Village Rule AGENTS.md does carry — claim → work → sign → release, "on completion write a RELEASE row and a signed ACK block" — which makes RELEASE the owner's act without inventing a clause.

Verified, per row

• register-read-path: register_status.py tracked, target resolves — I used it today to enumerate lanes, which is the capability the claim said was missing.
• cipher-build-pin: cipher-build-pin-check is wired as a prerequisite to up-core-capable and up-agents-stack, with a test.
• embedding-path: env::→dynamic:: landed. Disclosed, not implied: causes (2) ollama on 127.0.0.1 and (3) model not pulled are B850 node-local and cannot be measured from Z890. If still true, embeddings are degraded there even though the code lane is delivered — a node condition, not an open lane.
• lite-bootstrap: verified by blob hash, not ancestry — squash-merge makes the commits look unmerged. Also records that the branch's mode-120000 symlink to a SPARK-only absolute path correctly did not land, and fix(bootstrap): lite venv is CUDA-free — secrets paths stop pulling torch #2961 added the ignore rule itself.

Not backfilled, deliberately

SPARK's three other expired lanes (#2946, #2959, #2943) all have open PRs — active work past TTL, not abandoned work.

Two findings from doing this

1. register_status.py over-captures branch names from prose. B850's claim at line 2741 claims fix/cipher-mcp-transport-build-pin, and its scope says "Distinct from open PR #2864 (docs/reclaim-cipher-token-model)" — explicitly disclaiming that lane. The parser harvests backtick-quoted branch names from anywhere in the row, so it reports B850 as claiming the very lane its prose disclaims, manufacturing a phantom collision. Reported, not fixed — it's a real lane of its own.

2. The register carries 14 control characters beyond the known NUL: bare CR ×6, VT ×2, FF ×5. Python's splitlines() splits on all of them, \n counting does not, so the two disagree by 7 lines. That cost me a false accusation: I measured a 7-line offset and nearly reported the tool's line numbers as wrong. They're correct — my measurement was. Any line-based tooling on this file must count \n.

Expired-and-unreleased: 6 → 4

The remaining 4 are the phantom-collision row and SPARK's three live lanes.

Worth noting on the shape of this: HERMES filed the previous backfill (#2960, for Mavis) and then had its own lane expire unreleased three days later. That isn't carelessness — releasing is a separate manual act nobody's incentive points at, so it gets missed by the same agents who catch it in others. It's the argument for making release a gate rather than a habit.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• coderabbit

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: eb8e512b-913d-404b-ada9-338d0d50b81f

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}