feat(B-0400): bus-gate integration — allActiveClaims + --with-bus-claims (slice 5) by AceHack · Pull Request #2959 · Lucent-Financial-Group/Zeta

AceHack · 2026-05-13T06:52:48Z

Summary

claim.ts: export allActiveClaims() — collects active claims across all backlog items by discovering unique itemIds from claim-topic messages, then delegating per-item to activeClaims() for correct claim/release semantics
poll-pr-gate-batch.ts: --with-bus-claims flag appends active bus claims (busClaims?: ClaimRecord[]) to the BatchReport JSON, giving the autonomous loop PR-gate state + bus coordination state in one call
DST-clean injection: BusClaimsFn injectable type + pollFn injection in main() — tests never spawn gh or touch the filesystem

Slice context

Slices 1–4 shipped (types + CLI + watch + claim-coordinator + status subcommand). This is slice 5: the poll-pr-gate-batch.ts integration listed as deferred in the backlog item.

Usage after this slice:

bun tools/github/poll-pr-gate-batch.ts --with-bus-claims 2944 2953
# BatchReport now includes busClaims: [{ from: "otto", itemId: "B-0400", branch: "...", ... }]

Test results

bun test tools/bus/           → 64 pass, 0 fail   (+3 new allActiveClaims tests)
bun test tools/github/poll-pr-gate-batch.test.ts → 14 pass, 0 fail   (+4 new main() tests)
dotnet build -c Release       → 0 warnings, 0 errors
bunx tsc --noEmit             → 0 errors in changed files

Checklist

Rule 0: TypeScript only
Build gate: dotnet build -c Release → 0 warn / 0 err
Tests: 78 total pass, 0 fail
Bus claim acquired before work: claim.ts acquire --from otto --item B-0400
Backlog item updated with slice 5 scope
operative-authorization: aaron 2026-05-13: "cooling periods, harsh-critic reviewers, F# anchor sanity check"

🤖 Generated with Claude Code

…ims (slice 5) - claim.ts: export allActiveClaims() — collects active claims across all backlog items by discovering unique itemIds from claim-topic messages and delegating per-item to activeClaims() for correct release semantics - poll-pr-gate-batch.ts: --with-bus-claims flag appends active bus claims to the BatchReport output (busClaims?: ClaimRecord[]) - BusClaimsFn injectable type + pollFn injection in main() for full DST coverage without gh subprocess in tests - 3 new tests in claim.test.ts (allActiveClaims across items, with release) - 4 new tests in poll-pr-gate-batch.test.ts (flag present/absent, empty list) - bun test tools/bus/ → 64 pass; bun test poll-pr-gate-batch.test.ts → 14 pass - dotnet build -c Release → 0 warnings, 0 errors operative-authorization: aaron 2026-05-13: "cooling periods, harsh-critic reviewers, F# anchor sanity check" Co-Authored-By: Claude <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 57c0c7430c

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".

Copilot

Pull request overview

Adds bus-claim visibility to the PR gate batch poller so callers can retrieve PR gate state and cross-agent coordination state in a single JSON payload, aligned with B-0400 slice 5.

Changes:

Export allActiveClaims() from tools/bus/claim.ts to aggregate active claims across all backlog items.
Add --with-bus-claims to poll-pr-gate-batch.ts, appending busClaims to the BatchReport output (with injectable BusClaimsFn for deterministic tests).
Extend unit tests for both allActiveClaims() and main()/--with-bus-claims, and update the B-0400 backlog slice notes.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tools/github/poll-pr-gate-batch.ts	Adds `--with-bus-claims` output and injectable `BusClaimsFn`/`pollFn` to keep tests deterministic.
tools/github/poll-pr-gate-batch.test.ts	Adds `main()` coverage for `--with-bus-claims` and stdout capture.
tools/bus/claim.ts	Adds `allActiveClaims()` to aggregate active claims across items.
tools/bus/claim.test.ts	Adds tests validating `allActiveClaims()` behavior across multiple items and releases.
docs/backlog/P1/B-0400-inter-agent-ephemeral-communication-bus-nats-protocol.md	Documents slice 5 scope/details for the integration.

…ctiveClaims, test rename, doc count Three Codex/Copilot review findings resolved: 1. perf(claim.ts): refactor allActiveClaims() from O(items×messages) to single-pass. Previously: list() once to discover item IDs, then activeClaims(id) per item (each calling list() again). Now: one list() call, groupBy (from:itemId) in a single pass — identical tiebreaker logic to activeClaims(). 2. test(poll-pr-gate-batch): rename misleading test — was 'empty-PR early-return path' but actually tests 'busClaimsFn returns empty list'. Updated comment to clarify that --all-open empty path requires injectable listOpenPRs (not done here). 3. docs(B-0400): correct test count 4→3 in slice-5 scope section. All 33 claim tests + 14 poll-pr-gate-batch tests pass. Co-Authored-By: Claude <noreply@anthropic.com>

grok-4.3 collapses thinking/fast into one model identifier; the Mode parameter is preserved for future cursor-agent updates but is currently unread, causing TS6133 under noUnusedLocals. Co-Authored-By: Claude <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

+      timestamp: m.timestamp,
+      expiresAt: m.expiresAt,
+    });
+  }


…nfirmation 2026-05-13 (#3035) Aaron 2026-05-13 activated Otto on Claude Desktop alongside Otto-CLI, running the multi-foreground-surface architecture from PR #3030 for the first time. Preserves operational discoveries: 1. Claude Desktop Routines are GIT-TRACKED (durable across sessions) vs Claude Code CronCreate session-only — supersedes 'every session MUST CronList' mechanism at Desktop scope 2. Two cadence layers in Desktop: Loop (minute, session-bounded) + Routine (configurable, git-tracked). Both fire simultaneously. 3. CRITICAL ARCHITECTURAL GAP: SENDER_IDS in tools/bus/types.ts doesn't distinguish multi-surface instances. claim.ts treats --from otto from CLI and Desktop as IDENTICAL — split-brain prevention fails without schema extension. Vera caught this on PR #3032 review. Workarounds: lane-based + branch-prefix + future schema extension. 4. Approval friction profile shapes lane split natively: - Otto-CLI: better at .claude/rules/ edits (auto-accept covers) - Otto-Desktop: better at substrate + conversation + routines - The lane split emerges from approval costs, not convention — more robust because friction enforces it Generalizable principle: multi-foreground-surface works because surfaces have DIFFERENT cost profiles, not because they're redundant clones. Future multi-instance work should match work to lowest- friction surface for that work shape. Composes with: - Otto canonical bootstream (full canonical + tight) - .claude/rules/tick-must-never-stop.md (CLI cron caveat) - .claude/rules/claim-acquire-before-worktree-work.md (the rule with the architectural gap named) - .claude/rules/holding-without-named-dependency-is-standing-by-failure.md - B-0400 slice 3 (PR #2939) + slice 5 (PR #2959) - fetch-before-push memory (sibling discipline at git scope) Co-authored-by: Claude <noreply@anthropic.com>

…on for multi-Otto) (#3032) * docs(rules): claim acquire before worktree-creating backlog work (split-brain prevention for multi-Otto) Aaron 2026-05-13 set up Otto on Claude Desktop alongside Otto-CLI, creating a real multi-foreground-surface split-brain risk. Both Ottos share git + bus on one machine and could pick the same backlog row simultaneously. The claim-coordinator (tools/bus/claim.ts, B-0400 slice 3) was built for this — atomic check/acquire/release with PID-liveness + 24h TTL. PR #2939 shipped it; PR #2959 added the gate integration. The infrastructure exists; this rule is the discipline-level mechanization that auto-loads at cold-boot so both Ottos read it + use it before worktree-creating backlog work. Applies to: starting work on a B-NNNN row, creating feature branch + worktree, opening a backlog-advancing PR. Does NOT apply to: CI fixes on already-claimed PRs, ad-hoc memory- file writes, conversation-driven substrate, hot fixes. Composes with: - backlog-item-start-gate.md (this rule adds zero-th step) - dont-ask-permission.md (acquire exit 0 IS the substrate permission) - never-be-idle.md (if acquire fails, pick another row) - B-0400 + slice 3 + slice 5 infrastructure - Otto Claude Desktop bootstream (PR #3030) - fetch-before-push memory (sibling coordination pattern at git scope) Future substrate-level mechanization: pre-commit hook that auto-calls 'claim check' and fails on no-held-claim. Today's rule is the discipline-level fix. Co-Authored-By: Claude <noreply@anthropic.com> * fix(rules): correct Example 2 — multi-surface Otto can't be distinguished by same --from value (Vera P1) Vera caught (PR #3032 review): the original Example 2 showed two Ottos calling 'acquire --from otto' and the second exiting 1. That's WRONG per claim.ts line ~270 — same --from is filtered out as self-re- acquire (idempotent). Both calls succeed; split-brain not prevented. Fix: - Example 2 renamed to "KNOWN GAP" — shows actual behavior (both succeed) + names the architectural cause (SENDER_IDS doesn't distinguish multi-surface instances) - Added workarounds: lane-based convention, branch-prefix discipline, schema extension (otto-cli / otto-desktop) - New Example 3 shows TARGET behavior with schema fix (--from otto-cli vs --from otto-desktop) - Example 3 renamed to Example 4 (crash recovery) The architectural gap is a real follow-up: extending SENDER_IDS to support multi-surface instances of the same logical agent. Future work, not this PR. Operational consequence today: use the lane-based + branch-prefix workarounds. Claim-coordinator is INSUFFICIENT for multi-Otto without the schema extension. Co-Authored-By: Claude <noreply@anthropic.com> * docs(rules): fix false split-brain + PID-liveness claims P1 (PRRT_kwDOSF9kNM6B5VwU): carved sentence claimed split-brain is but claim.ts filters by c.from !== sender, soprevented same-sender instances both exit 0. Updated to state --from must differ and document the known gap explicitly. P2 (PRRT_kwDOSF9kNM6B5VwX): example 4 claimed PID-liveness can reclaim a claim mid- no such mechanism exists. Only TTLTTL expiry or explicit release ends a claim. Corrected comment. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(rules): remove branch-prefix as split-brain workaround (Vera P2 caught it doesn't work) Vera caught: branch-prefix doesn't prevent split-brain because claim.ts only filters by 'from', not by 'branch'. Two Ottos with same --from otto but different --branch both succeed (both exit 0). The branch field is post-hoc metadata, not a coordination key. Real workarounds reduce to: 1. Lane-based convention (zero-code; ONLY real prevention today) 2. Schema extension (future SENDER_IDS additions) Co-Authored-By: Claude <noreply@anthropic.com> * fix(rules): carved sentence consistency — remove branch-prefix from workarounds (Vera P2) The body of the rule was already corrected (commit 030513f) to remove branch-prefix as a workaround. The carved sentence wasn't updated. Vera caught the inconsistency. Now carved sentence + body both say: lane-based convention is the ONLY real split-brain workaround today. Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings May 13, 2026 06:52

AceHack enabled auto-merge (squash) May 13, 2026 06:52

Copilot started reviewing on behalf of AceHack May 13, 2026 06:53 View session

chatgpt-codex-connector Bot reviewed May 13, 2026

View reviewed changes

Comment thread tools/bus/claim.ts Outdated

Copilot AI reviewed May 13, 2026

View reviewed changes

Comment thread tools/bus/claim.ts Outdated

Comment thread tools/github/poll-pr-gate-batch.test.ts Outdated

Comment thread docs/backlog/P1/B-0400-inter-agent-ephemeral-communication-bus-nats-protocol.md Outdated

AceHack and others added 2 commits May 13, 2026 03:00

Copilot AI review requested due to automatic review settings May 13, 2026 07:07

Copilot started reviewing on behalf of AceHack May 13, 2026 07:07 View session

AceHack merged commit 7c622df into main May 13, 2026
29 of 30 checks passed

AceHack deleted the feat/b-0400-slice5-bus-gate-integration branch May 13, 2026 07:11

Copilot AI reviewed May 13, 2026

View reviewed changes

Comment thread tools/bus/claim.ts

timestamp: m.timestamp,

expiresAt: m.expiresAt,

});

}

This was referenced May 13, 2026

feat(B-0400): multi-agent review close — Otto protocol review + acceptance-criteria close (slice 6) #2969

Merged

docs(rules): claim acquire before worktree work (split-brain prevention for multi-Otto) #3032

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(B-0400): bus-gate integration — allActiveClaims + --with-bus-claims (slice 5)#2959

feat(B-0400): bus-gate integration — allActiveClaims + --with-bus-claims (slice 5)#2959
AceHack merged 3 commits into
mainfrom
feat/b-0400-slice5-bus-gate-integration

AceHack commented May 13, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

AceHack commented May 13, 2026

Summary

Slice context

Test results

Checklist

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants