Repository navigation
docs: prune the shipped entries from TODOS.md - #406
Conversation
Removes 21 entries whose work has shipped or whose code is gone, each checked against git history, merged PRs, closed issues and the current code, plus the empty P3 CLI heading. Entries that source comments or docs still cite by title stay until those comments move. The P1 entry "Finish removing base machines" is unchanged. The tab-hover preview entry now links its issue, #137. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
🟡 Changes recommended
Several retained TODOs now contain stale instructions or dangling references to deleted entries.
3 open findings
What changed in this PR
Prunes shipped or obsolete work from TODOS.md while preserving active entries.
Changes:
- Removes 21 completed or obsolete TODO entries.
- Removes the empty P3 CLI section.
- Links the tab-hover preview entry to issue #137.
| File | Description |
|---|---|
TODOS.md |
Removes completed entries and adds an issue reference. |
🧠 Review effort: Balanced
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
The foreground-process entry no longer points at the deleted SessionPTY and SessionDaemon, the eyeball-only snapshot entries no longer count or point to the removed #136 entry, and the WORKROOM-2T note names the fingerprint code instead of a removed section. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ae951c9408
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".
| opens the window is rare. Cheap to close, and the failure it produces (a silently lost entry, no | ||
| error) is one of the hardest to diagnose from a bug report. | ||
|
|
||
| ## Recently done |
There was a problem hiding this comment.
When a user adds a moved or broken linked worktree whose .git file still begins with gitdir: but points to a nonexistent target, this work remains unfinished: internal/vcs/vcs.go:50-58 checks only the prefix, and internal/vcs/vcs_test.go:577-580 explicitly accepts gitdir: /elsewhere. Removing the TODO therefore falsely records its validated-pointer requirement as shipped and loses tracking for the resulting late failure during workroom operations; retain or narrow the entry until the referenced git directory is actually validated.
AGENTS.md reference: AGENTS.md:L44-L50
Useful? React with 👍 / 👎.
Summary
PR 4 of the docs and repo hygiene lane of the architecture cleanup plan (phase 1: prune
TODOS.md).TODOS.mdhad 104 entries. Each one was checked against git history, merged PRs, closed issues and the current code (not against the entry's own FIXED or SHIPPED label). 21 entries whose work has shipped, or whose code is gone, are removed; the file goes from 331K to 254K. Full write-ups of the removed entries stay in git history. The P1 entry "Finish removing base machines" is kept untouched. The emptyP3 — CLIheading went with its two entries. TheRecently donesection is untouched.Kept although shipped (11). Source comments, tests,
macapp/project.ymlormacapp/AGENTS.mdstill cite these entries by title. The Swift and YAML comments belong to other lanes, so these stay until those comments move: "Bump the libghostty pin"; "Bundled ghostty resources — provenance recorded, nothing to regenerate"; ".task(id:)cancellation is not reliably delivered on an in-place value swap — measured 2"; "N-in-flight concurrency accounting test for status sweeps — Muxy test-practices review fol"; "Repo-level meta-test for Makefile/CI test invariants"; "Git diff shows one side when a file is both staged and re-modified"; "Profile and fix the theme picker's invalidation storm"; "Stopped run-tab silently closes instead of warning when its command is cleared — #127 foll"; "Status-aware avatar image loader"; "FilesPanelrenders up to 4000 rows eagerly"; "Focus-responder App Hang on a large diff — WORKROOM-2T".Partly shipped, kept whole (7). "Credential broker clients: the review's deferred items — #251 review follow-up"; "Bump the libghostty pin again — package 1.5.20260903 / ghostty
c4e16970"; "VCS write actions — Phase 2 — roadmap pointer"; "Agent-routed VCS writes: the findings the/reviewpass didn't fix — issue #205"; "The ✦ diagnosis badge was unreachable to accessibility"; "Consolidate terminal focus authority — focus-desync follow-up"; "WindowDragUITests.testDraggingWorkroomTabReordersTwoChipsis a known gesture-timing flak". Trimming them to the open remainder is a judgement call left for you.Issue links. Only one remaining entry has an issue of its own ("Tab-hover live mini-preview", #137, closed), so it now carries
**Issue:** #137. Most entries cite a closed parent issue they follow up, which is not an issue for the work itself. I have not filed anything: the 82 remaining entries with no issue of their own are listed below for you to decide.Removed entries and the evidence
PendingMachines.add(wri...workroom-sessiondaemon: the findings the/reviewpass didn't fix — persist-sessions fselection_changednotificationcase GHOSTTY_ACTION_SELECTION_CHANGED, GhosttySurfaceView.swift:1212func handleSelectionChanged()with `...withTimeoutdoesn't observe the CALLER's own cancellation/reviewpass didn't fixCommandResultconsumers don't know aboutsignaled— gh-flap eng-review follocase interrupted, :149 `if result.signaled...VCSToolVersionCachepins a tool verdict for the whole process — gh-flap follow-upbelowFloorTTLto VCSToolVersions.swift). Code: VCSToolVersions.swift:139actor VCSToolVersionCache, :162 `init(...workingStatusonto theVCSProvidingprotocol — VCS-foundation follow-upflagsChangedthrew an ObjC exception on every modifier pressstatic func unshiftedCodepoint(for:)with `guard event.type == .keyDown // event.type == .keyUp...ChangedFileRowInvalidationTestsflaked only inside the full serialapp-testrunSessionByteQueue'swouldBlockpath has no dedicated testvcs.Detectto validate a real repo — #103 follow-upwithLockauditRemaining entries with no issue of their own
Say which of these should become GitHub issues (all, none, or a list) and I will file them with the
needs-triagelabel and link them here.c4e16970/reviewpass didn't fix — issue Agent-backed VCS writes #205 (follows up Agent-backed VCS writes #205)WindowDragUITests.testDraggingWorkroomTabReordersTwoChipsis a known gesture-timing flakwaitUntilExitnever returned for an exited agent — fix: the #220 low findings #223acceptpath is untested — feat(macapp): forward a loopback port over the agent multiplex (#208, app half) #218 review follow-upworkroom-sessionshim — issue Remote agent sessions #154 follow-up (follows up Remote agent sessions #154)ghostty_surface_tty_nameis available and unused — filed 2026-08-14AppTerminalViewfor action-dispatch plumbing — reviewed 2026-08-12, noAgentVCSConnectiononce it carries files and events — Agent-backed file service: listing, raw reads and change notification #211 follow-up (follows up Agent-backed file service: listing, raw reads and change notification #211)vcs.rs— Agent-backed file service: listing, raw reads and change notification #211 follow-up (follows up Agent-backed file service: listing, raw reads and change notification #211)XDG_CONFIG_HOMEnever reaches theghauth probe — gh-flap eng-review follow-up (follows up PR sidebar: "Github CLI not signed in" #86)ShellEnvironmentexposes no probe-readiness signal — gh-flap eng-review follow-up (follows up PR sidebar: "Github CLI not signed in" #86)FileDiffhunks) for the diff viewer — 47b / VCS-foundation follow.io— error-taxonomy follow-upDefaultskeys with dots in their names cannot be observed — measured, NOT fixed (2026-09terminfo.versioninto the resource-contract test — filed 2026-09-03~/.configthemes into families — Theming #36 follow-up (follows up Theming #36)theme = dark:<X>,light:<Y>syntaxminimum-contrastprecedence vs a user's own ghostty configAppStore's defaultcliisn't hardened against a forgotten test double — filed 2026-08-.task(id:)cancellation is not reliably delivered on an in-place value swap — measured 2FilesPanelrenders up to 4000 rows eagerlyPre-Landing Review
Docs only. Verdicts came from five read-only passes over the file, each citing a commit, a merged PR, a closed issue or a code check per removed entry. Known gaps: two entries' recurrence claims could not be verified (Sentry WORKROOM-3A, the ChangedFileRow flake) and are kept; the
OSC 99 desktop notificationsentry is stale (ghostty PR #10467 closed unmerged) but kept because the upstream issue is still open.Test plan
TODOS.mdcites a removed entry by title (the cited ones are kept)🤖 Generated with Claude Code