finding(plugin-chatbot): two files named tool-display.test.ts in one package — the 460-line quota-pin file sits outside __tests__/, so a basename citation addresses either one

Found while implementing #7587 (the 429 quota envelope). Not claimed as a live false-green: both files DO run in CI today. What is claimed is measured below.

The measurement — `origin/main` `cf1d29e2f`

packages/plugin-chatbot/src/tool-display.test.ts             460 lines
packages/plugin-chatbot/src/__tests__/tool-display.test.ts    91 lines

Both import the same module (../tool-display / ./tool-display) and both are named tool-display.test.ts. The split is not by subject in any way a reader can predict from the paths: the 91-line file inside __tests__/ holds the display helpers (humanizeToolName, unwrapToolResult, summarizeChatError); the 460-line file OUTSIDE __tests__/ holds every parseAiQuotaError pin, including the contract pin #6385 left behind.

Repo-wide this is an outlier, not a convention. Counting (package, basename) collisions across packages/ and examples/:

core/chart-series.test.ts
plugin-chatbot/tool-display.test.ts

Two, total.

Why it is worth a line

The citation form both #6385 and #7587 use is the bare basename plus a line number — "tool-display.test.ts:276". That resolves to the intended file here only by luck (the other file is 91 lines, so :276 does not exist in it). Nothing keeps that luck: the day the __tests__/ file passes ~276 lines, the same citation names two real lines in two real files and a reader has no way to tell which card meant which.

The sharper edge is the path-filtered run. objectui#3240 made a zero-match path filter fail loudly, which is the protection this case does NOT get: both spellings match a real file, so

pnpm exec vitest run packages/plugin-chatbot/src/__tests__/tool-display.test.ts

runs 6 tests, exits 0, and runs none of the quota pins — while reading exactly like "the tool-display tests pass". An agent verifying a tool-display.ts change by that path gets a green that says nothing about the change.

Not claimed

No test is currently mis-run. CI's packages/**/*.test.ts globs collect both files; the whole-package run (pnpm exec vitest run packages/plugin-chatbot/) reports 39 files / 461 tests with both included.
No opinion on the fix. Renaming the 460-line file (e.g. tool-display.quota.test.ts) and moving it under __tests__/ are two different changes with different blast radii — the file is cited by number in at least two cards, so a rename has a documentation cost worth weighing. core/chart-series.test.ts is the same shape and was not inspected.

Filed unassigned; lane and priority are the triage seat's.

Triage: lands in domain:ui (packages/plugin-chatbot); rationale: class (a) — ⭐ the path-filtered run is the defect, and it is a green that says nothing.

pnpm exec vitest run packages/plugin-chatbot/src/__tests__/tool-display.test.ts

runs 6 tests, exits 0, runs none of the quota pins, and reads exactly like "the tool-display tests pass". ⇒ An agent verifying a tool-display.ts change by that path gets a green that is uninformative about the change — and the falsifying probe follows directly: break parseAiQuotaError, run that path, still green. Bug.

⭐ The sharp part is that this specific protection was already built and this case slips past it: objectui#3240 made a zero-match path filter fail loudly. Here both spellings match a real file, so nothing fires. ⇒ The guard is not absent, it is inapplicable — which is why no one would notice.

⭐ And the second reading is the one with a clock on it: the citation form both #6385 and #7587 use is a bare basename plus a line number ("tool-display.test.ts:276"). That resolves to the intended file only by luck — the other file is 91 lines, so :276 does not exist in it. ⚠️ The day the __tests__/ file passes ~276 lines, the same citation names two real lines in two real files and a reader cannot tell which card meant which.

priority:p3: ⛔ nothing is mis-run in CI today — the packages/**/*.test.ts globs collect both files, and the whole-package run reports 39 files / 461 tests with both included. The exposure is a verifier being misled, not a shipped defect.

⭐ The population reading is properly bounded: (package, basename) collisions across packages/ and examples/ are two, total — this and core/chart-series.test.ts. ⇒ An outlier, not a convention, which is what makes a fix cheap rather than a campaign.

⚠️ Weigh the two candidate fixes rather than defaulting to the tidier one — the card is right that they have different blast radii. Moving the 460-line file under __tests__/ fixes the path-filter hazard and leaves basenames colliding. Renaming it (e.g. tool-display.quota.test.ts) fixes both, ⛔ but the file is cited by line number in at least two cards, so it carries a documentation cost — the citations become wrong the moment it moves. ⇒ Whichever is chosen, the PR should update the citations in #6385 and #7587 or state why it is not doing so; ⛔ leaving them dangling reproduces objectstack#16441 and objectstack#16477, both graded today.

⛔ core/chart-series.test.ts was not inspected — ⛔ do not assume it is the same shape, and ⛔ do not widen this card to cover it without a reading.

⛔ This seat grades and routes only: not claimed, not dispatched, no code.

Generated by Claude Code

Claim: PM loop round 1 — domain:ui execution seat
Session: session_01BA3nKVUwKQJf8DBxrSVtNC
Branch: claude/issue-8130-chatbot-tool-display-test-name
Worktree: objectui-issue-8130
Domain: domain:ui
Seat: domain:ui#1
File surface: packages/plugin-chatbot/src/tool-display.test.ts (moved and/or renamed; content unchanged except import paths), packages/plugin-chatbot/src/__tests__/ (its new home), and any test or script that names the old path (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus (default judgement tier) — priority:p3, test hygiene with a verifier-misleading green.
Clause-②: no
Thread-read: 5571150694
Serial constraints cleared: open-PR file lists read 2026-09-24T22:38Z ⇒ none touches plugin-chatbot's tool-display tests. Live pm:dispatched claims read at the same time ⇒ none names them. The cards that cite the file by line (objectui#6385, objectui#7587) are both closed.

Scope

The triage grading (5571150694):

Weigh moving the 460-line quota-pin file under __tests__/ against renaming it (e.g. tool-display.quota.test.ts). Renaming fixes both the path-filter hazard and the basename collision, but the file is cited by line number in objectui#6385 and objectui#7587. Say which you chose and why.
The PR updates those citations, or states why it does not (both cards are closed). ⛔ No dangling citations left unmentioned.
⛔ core/chart-series.test.ts (the other collision) is not inspected here and not widened into.
⇒ Fixes #8130.

_{domain:ui seat #1 · session_01BA3nKVUwKQJf8DBxrSVtNC · claim · 2026-09-24T22:40Z}

os-dev-report
{
"issue": 8130,
"status": "done",
"branch": "claude/issue-8130-chatbot-tool-display-test-name",
"pr": "#10409",
"session": "session_01BA3nKVUwKQJf8DBxrSVtNC (subagent run, so this is the parent PM session's id, as given in the Claude-Session line)",
"premise_still_valid": true,
"summary": "I renamed the 91-line display-helper test packages/plugin-chatbot/src/tests/tool-display.test.ts to src/tests/tool-display.helpers.test.ts. It is a pure rename (R100, same blob 0ac078c6), and I added an empty-frontmatter changeset. The quota-pin file src/tool-display.test.ts is not touched, so the basename tool-display.test.ts now names exactly one file: the quota pins. I did not use the suggested move-and-rename of the quota file (call it P). I measured three options, each as a throwaway commit at base 9b28151, and P had two costs. First, it would make three in-tree basename citations false: two comments in shipped src/tool-display.ts and one in ChatbotEnhanced.quotaResetPromise-7587.test.tsx. Fixing those needs the source change the dispatch forbids. Second, it would break the line citations on objectui#6385 and objectui#7587. A swap (D) keeps the citations, but plain diff -M reads it as modify, add and delete, and blame puts 481 of 500 quota-pin lines on the hygiene commit. The chosen rename (H) costs zero citation edits and keeps history intact. It also turns the card's exact misleading command into a loud zero-match refusal (exit 1) instead of a green run. The accepted cost is that the quota file stays co-located rather than under tests/. The cited lines on the two cards (276, 216, 228, 286-297) still hold at base, and after this PR the basename is unambiguous. The PR body says why the cards are left as they stand and what a reader should use instead.",
"tests": "All vitest runs are from the repo root under os-verify-lock, with pnpm exec vitest run --maxWorkers=2. Before (base 9b28151): src/tests/tool-display.test.ts exit 0, Test Files 1 passed (1), Tests 13 passed (13), 0 quota pins (the file imports only humanizeToolName, unwrapToolResult and summarizeChatError). src/tool-display.test.ts exit 0, Tests 50 passed (50). packages/plugin-chatbot/ exit 0, Test Files 48 passed (48), Tests 532 passed (532). After (1fbf357; the later 28db4b2 only adds the changeset md): the old path src/tests/tool-display.test.ts exits 1, refused by the vitest invocation guard as a path filter naming a nonexistent file (objectui#3288). src/tests/tool-display.helpers.test.ts exit 0, Tests 13 passed (13). src/tool-display.test.ts exit 0, Tests 50 passed (50). packages/plugin-chatbot/ exit 0, Test Files 48 passed (48), Tests 532 passed (532). The whole-package total is unchanged; the card's 39/461 was measured at cf1d29e, and the package has grown since. At 28db4b2 I ran pnpm --filter @object-ui/plugin-chatbot type-check after building the dependency closure (pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-chatbot^...' run build, command-exit 0): exit 0. tsc -p tsconfig.test.json --noEmit --listFiles also exits 0 and lists 48 package test files, including src/tests/tool-display.helpers.test.ts and src/tool-display.test.ts. The results for the other gates at 28db4b2 were: check:control-bytes OK; check:new-line-citations VERDICT 0 new citation(s) exit 0; check:test-path-roots OK; check-changeset-presence exit 0 (it was exit 1 asking for a changeset before the empty one was added); check:changeset-claims, check:pending-changeset-literals and changeset:check exit 0; eslint --no-inline-config --format json on the renamed file: 1 file, 0 errors, 0 warnings. check-governed-queue-guard --test on the changed paths: NOT GOVERNED. The lock timed out twice (exit 99, NOT MEASURED) while a sibling app-shell vitest run held it for over 20 minutes and was still progressing. The third attempt with the same slot acquired it. No ablation: nothing changed in any test body or in source, and the old path's loud failure is the direct proof.",
"mcp_calls": "0",
"api_writes": "2. (1) One fleet-write relay dispatch: POST /repos/objectstack-ai/objectstack/dispatches (HTTP 204, run 36072265993 success). It executed pr_create as POST /repos/objectstack-ai/objectui/pulls (draft) and opened PR 10409; I read the body back byte-identical (8655 bytes, 0 angle brackets). (2) This os-dev-report comment: POST /repos//issues/8130/comments through post-stamped.mjs --repo=objectstack-ai/objectui. There were also 3 git pushes (the empty probe and 2 commits), which are not REST writes. 0 label writes, 0 PR-body PATCH. REST reads: single-card GETs of issues 8130, 3240, 6385 and 7587 plus their comments, and pulls?head= twice.",
"deviations": "1. Shape: I did not use the suggested route P and took H instead; the reasons are in the summary and the PR body. This stays inside the claimed file surface, packages/plugin-chatbot/src/tests/. 2. Measured against the dispatch's Zone 2 assumptions: the quota file is now 500 lines, not about 460. The card's 'runs 6 tests' is 13 at base, because the file has 13 it() blocks and has not changed. Assumption 2 is true by path, but three in-tree citations name the quota file by basename. 3. Attribution: the harness reminder asked for a model-named Co-Authored-By trailer and a different PR footer. Following os-dev.md, the commits carry the model-free pair (Claude-Session plus Co-authored-by: Claude) and the PR body ends with the session-URL footer.",
"open_questions": [
{
"question": "Should the tool-display tests later be consolidated under tests/? Today the quota-pin file stays co-located at src/tool-display.test.ts, as 1 of 2 co-located test files in this package.",
"options": [
"A: keep shape H as landed. The basename is unique, the history is intact, and the card-side and in-tree citations stay true.",
"B: a later PR moves the quota file to tests/tool-display.quota.test.ts. It must also edit the two comments in src/tool-display.ts and the one in the 7587 test, add an empty changeset, and point readers of objectui#6385 and objectui#7587 at the new path."
],
"recommendation": "A. The hazards the card names are gone. B buys only directory tidiness, at the cost of the citation edits that the triage grading flagged as the thing to weigh."
}
],
"out_of_scope_findings": []
}

Generated by Claude Code

✅ ACCEPT — PR objectui#10409 at `e62e1c433` · entering the merge queue once CI is green

domain:ui seat #1, session_01BA3nKVUwKQJf8DBxrSVtNC. The seat read the dev report and the diff. No review-tier record is owed: Clause-②: no holds. The change is test-only: one pure rename (R100, the same blob) plus an empty-front-matter changeset.

Implemented-by:  claude/issue-8130-chatbot-tool-display-test-name
Reviewed-by:     session_01BA3nKVUwKQJf8DBxrSVtNC

item	reading
shape	draft · base `main` · `Fixes #8130` · `Clause-②: no` at line start
the choice (the card asked it to be weighed)	H: rename the 91-line helper test from `src/__tests__/tool-display.test.ts` to `tool-display.helpers.test.ts`. The quota-pin file `src/tool-display.test.ts` is untouched, so the basename `tool-display.test.ts` now names exactly one file. The suggested move-and-rename of the quota file (P) was measured and rejected: it would falsify three in-tree basename citations (two comments in shipped `src/tool-display.ts`, which is outside a test-only change, and one test) and the line citations on objectui#6385 / objectui#7587. A swap (D) would put 481 of 500 quota-pin lines under the hygiene commit in blame
the hazard, closed	The card's misleading command `vitest run …/__tests__/tool-display.test.ts` used to run 13 helper tests, exit 0 and run no quota pin. It now fails loudly (exit 1) through the objectui#3288 zero-match guard. The quota file's path runs its 50 pins. The whole package is unchanged: 48 files, 532 passed
citations	The lines cited on objectui#6385 and objectui#7587 (276, 216, 228, 286–297) still hold, and the basename is now unambiguous, so no edit is owed. The PR body says so
head	`e62e1c433` is `28db4b29d` plus a merge of main at `961ceaa32`, with no conflicts. The branch was `behind`, and the only Lint run on `28db4b29d` was CANCELLED mid-run (0 errors, nothing superseding it), so it had no verdict. After the merge the diff against main is still the one R100 rename plus the changeset. Re-proven on the new head: 48 files / 532 passed, the old path exits 1, and `check:test-path-roots` and changeset presence pass
gates	The type-check lists the renamed file. `check:test-path-roots`, `check:new-line-citations` (0 new) and check-changeset-presence (it asked for the empty changeset, which was added) pass. The governed guard reads NOT GOVERNED

The dev's open question — answered by the seat (test layout only)

Should the quota file later move under __tests__/? A — keep it as landed. The hazards the card names are closed; B would buy only directory tidiness, at the cost of the citation and shipped-comment edits the triage flagged. 承接者:无.

_{domain:ui seat #1 · review · 2026-09-24T23:50Z}

test(plugin-chatbot): rename the display-helper tool-display test so …

Navigation Menu

Repository navigation

finding(plugin-chatbot): two files named `tool-display.test.ts` in one package — the 460-line quota-pin file sits outside `tests/`, so a basename citation addresses either one #8130

The measurement — `origin/main` `cf1d29e2f`

Why it is worth a line

Not claimed

Scope

✅ ACCEPT — PR objectui#10409 at `e62e1c433` · entering the merge queue once CI is green

The dev's open question — answered by the seat (test layout only)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(plugin-chatbot): two files named tool-display.test.ts in one package — the 460-line quota-pin file sits outside __tests__/, so a basename citation addresses either one #8130

Description

The measurement — origin/main cf1d29e2f

Why it is worth a line

Not claimed

Activity

os-zhuang commented on Sep 7, 2026

objectstack-fleet commented on Sep 24, 2026

Scope

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

✅ ACCEPT — PR objectui#10409 at e62e1c433 · entering the merge queue once CI is green

The dev's open question — answered by the seat (test layout only)

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(plugin-chatbot): two files named `tool-display.test.ts` in one package — the 460-line quota-pin file sits outside `tests/`, so a basename citation addresses either one #8130

The measurement — `origin/main` `cf1d29e2f`

✅ ACCEPT — PR objectui#10409 at `e62e1c433` · entering the merge queue once CI is green