stack/agents/filbert/work/queue-142/r1/body.md at dev-console

0 Fork 1

Files

T

jason.woltje b9ea9304fd lead: commit filbert's untracked evidence (273 files: queue-101 queue-102 queue-105 queue-107 queue-108 queue-109 queue-110 queue-112 queue-113 queue-114 queue-115 queue-116 queue-119 queue-120 queue-122 queue-123 queue-125 queue-126 queue-128 queue-129 queue-130 queue-131 queue-132 queue-133 queue-134 queue-135 queue-136 queue-137 queue-139 queue-140 queue-141 queue-142 queue-145 queue-146 queue-147 queue-148 queue-149 queue-150 queue-151 queue-152 queue-153 queue-154 queue-155 queue-156 queue-162 queue-163 queue-165 queue-166 queue-167 queue-169 queue-170 queue-175 queue-177 queue-181 queue-182 queue-184 queue-185 queue-186 queue-187 queue-188 queue-49 queue-64 queue-66 queue-70 queue-71 queue-80 queue-84 queue-85 queue-87 queue-88)

Unclaimed files left in the canonical checkout. Inventory: canonical-cleanup-inventory.md (commit-as-evidence groups). detect-secrets scan clean on 2026-11-02: hits are sha256 pins, git object ids, test sentinels already in tracked tests and unit-file lines.

2026-11-02 15:05:14 -06:00

4.2 KiB

Raw Permalink Blame History

Row 142 round 1 review (Filbert): approve

Candidate: manifest sha256 1220d1341f35527ebd9c2b8cc86fa049617a19a751f6e72b87e5c483517348e7, 4 paths, base 8e2c5c2d0, packet agents/researcher/work/queue-142/ in df5bbfb82. The paths are cli.mjs, verbs.test.mjs, docs/TOOLS.md and packages/cli/README.md. The README is in by Sage's ruling (sage-0388).

Rebuild

A fresh worktree at refactor HEAD with row142.patch applied. Every manifest hash matched and git diff --check was clean.
The added lines have no dash, curly-quote or control characters. The packet manifest checks.
No commit from 8e2c5c2d0 to HEAD touches a manifest path.

What I checked and found correct

The filter. if (!follow && m.in_reply_to !== sent.id) continue; sits before the print and before message.read, so a skipped message is neither printed nor read.
- !follow implies receiveOnly is false, so sent is always set where sent.id is read.
- Under --follow and follow the line is inert, so both behave as before.
What a skipped message becomes. message.receive marks each row it returns as delivered and never returns a delivered row again (broker.mjs:1258). A skipped message is therefore delivered but unread.
- mosaic inbox pages messages, keeps those without read, and prints each in full with formatMessage.
- So "leaves the rest unread for mosaic inbox" holds, and the new timeout text ("mosaic inbox shows it") is true.
The human's unread messages block nothing. The thread release check counts only messages to the thread instance.
Tests against the brief.
- Test 1 has another role's note (cto) and the same role's unsolicited note (coder, no in_reply_to) arrive before the reply. It asserts that neither is printed, that there is exactly one message.read, and that mosaic inbox lists both in full as 2 unread.
- The same-role case matters. It catches a filter on the sender instead of in_reply_to (my mutant B below).
- --wait 5 turns a never-printed reply into an assertion failure, not a hang.
- The late-reply test now asserts the next talk skips it, reads nothing, and inbox lists it. The Console replies assertion is kept.
The docs. The TOOLS.md and README text match the code.

Notes (non-blocking)

Skipped messages can't be marked read from the CLI.
- The only message.read caller in cli and webui is talk's loop.
- follow receives only undelivered messages, so it won't print them either. Researcher's build.md discloses this.
- So mosaic inbox's unread list only grows. A follow-up could add a way to mark them read, for example on inbox.
The Console's send-section text is now false for users. packages/webui/src/public/bus.js:280 says the next mosaic talk in a terminal still prints each reply, and :261 says the same in a comment.
- send-browser.test.mjs:105 asserts that text. broker.mjs:1519 has the same comment.
- These are outside the row's files, and public/ needs decision 123. They are on Researcher's follow-up list, and I'd queue the Console text soon, since users see it.
One TOOLS.md line is longer than the paragraph's wrap (the line ending "stop sends SIGTERM to a run's runner after checking it in"). Cosmetic, and doc-only under decision 137.

What I ran and what I took from receipts

I ran, as single files on the candidate with TMPDIR set and Docker disabled:

Test file	Result
cli verbs	6/6
cli follow	6/6
cli tester	6/6

I also ran two mutants of my own on cli.mjs, one at a time against verbs.test.mjs. The tree matched the manifest again afterwards.

Mutant	verbs
A: skip after the print (every message printed, only the reply read)	2 fail
B: filter on `from_role !== to` instead of `in_reply_to`	2 fail

From Researcher's receipts, not rerun by me:

m1 to m6, all killed by assertion, with 3 controls passing (mutants-run.txt). This includes their disclosed correction: m3 first ended by timeout and was rerun after --wait 5.
The gate from 22:39:44Z to 22:43:18Z, all rc=0 (suites-run.txt): cli 206, business 101, bus 214, plugins 21, tasks 97, webui 113, harness 90, seat 135, and the ten scripts/test-*.sh.

I took no slot.

4.2 KiB Raw Permalink Blame History