diu word lists: five kinds of wording to avoid in replies

[EdbertChan]

Summary

Five new word lists say which kinds of wording confuse the user in replies, with real examples of each.

They come from twenty times the user asked "what do you mean" in recent chats.

They live with diu, the reply checker. Nothing reads them yet; the PR above this one has diu use them.

Review Claim

Five phrase lists, engine/hooks/diu-stop/phrases/plain-words-*.json, one per kind of confusing wording, each loadable by phrases.load(name, directory=...), plus a test that they load and carry no PR numbers or dates.

Review Lane

behavior

Review Unit

engine-runtime

Safety Invariant

Data files and one test only. Nothing reads these lists yet, so no hook and no reply changes.

Slice Rationale

The lists are what diu will read, so they land first and on their own, in diu's own folder.

The branch is now one commit on main. It previously sat on the phrase-list-format PR, which has merged, and carried that PR's commits too; those are gone, so the diff is these six files.

The five kinds are: words the assistant made up, raw code or setting names, internal tool names, tech jargon and abbreviations, and status words. Each list reads the user's message as well as the reply, so a word the user used first does not count.

Non-goals

• diu does not read the lists in this PR.
• No change to the phrase-list loader or the judge.
• No regex: the model judges meaning against the examples.

Test Plan

Test Plan

Run on the rebuilt branch, one commit on main:

• python3 -m unittest discover -s engine/hooks/diu-stop/tests: Ran 97 tests ... OK
• uvx ruff check . --select E9,F: All checks passed!
• The rebuild keeps exactly the old changes: git patch-id --stable matches the old range's patch id
• Live check with the real judge, Claude runner only, loading plain-words-made-up-labels from diu's folder: "No hook decides differently. The hooks only gain replay functions." returned match: True, closest "The hooks only gain replay functions." (7.6s); "Yes. The hook's logic stays the same; this is a simple refactor." returned match: False (9.4s).

Revert Plan

Revert Plan

• Safe to revert? Yes. Nothing reads the lists.
• Revert command: git revert <merge-sha>
• Post-revert steps: None.
• Data migration? No

🤖 Generated with Claude Code

https://claude.ai/code/session_01Gnua5cuAAF1ey1b7zRqJCC

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_465c380a-e14e-4dd4-89e5-3cb519bae037)

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_b460767c-a672-4ac1-82d6-39e3e6ccc911)

[mergify]

Queued — the merge queue status continues in this comment ↓.

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_c606b307-3470-441f-8724-6fb26589ab43)

[EdbertChan]

@Mergifyio queue

[mergify]

Merge Queue Status

• ✅ Entered queue — 2026-09-12 01:15 UTC · Rule: admin-bypass · triggered by @EdbertChan with the @mergifyio queue command
• ✅ Checks passed · on draft merge queue: checking #446 on main (94593f3) #459
• ✅ Merged — 2026-09-12 01:20 UTC · at a3d24a852efad69f9c2f5b2ea31c7b839c3a72de · squash

This pull request spent 5 minutes 23 seconds in the queue, including 4 minutes 4 seconds running CI.

Required conditions to merge

• check-success = lint
• check-success = test