Skip to content

feat: create a folder from the project picker's browse tab - #31644

Merged
bolichen97 merged 1 commit into
mainfrom
fix/26005-project-picker-new-folder
Nov 7, 2026
Merged

bolichen97 merged 1 commit into
mainfrom
fix/26005-project-picker-new-folder

Conversation

@bolichen97

@bolichen97 bolichen97 commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Let people make a new folder from the project picker's Browse tab, so starting a project in a folder that does not exist yet no longer needs a terminal.

The Browse tab lists the gateway host's folders and Selects one as the session's project, but it cannot make one. To start in a new folder, a user leaves the dashboard, runs mkdir in a shell on the gateway host, comes back and browses to it.

Builds on #26007 by @alparamonov.

Why it matters

Starting a project in a new folder is a common first step, and today it costs a switch to a shell on the gateway host for a one-word action. On a remote gateway that shell is on a different machine from the user's own.

Not a goal

  • Creating several folder levels (a/b/c) in one step, or creating a folder from a typed path that does not exist.
  • Renaming, moving or deleting folders.
  • A create control in the activity-panel file browser or the workspace picker.
  • A keyboard shortcut for the control. (The one keyboard change here, Tab on Browse, was a maintainer ruling during review.)

What changed (motivation → approach → change)

Starting a project in a folder that does not exist yet needed a shell on the gateway host, so the picker now makes the folder itself. The Browse tab gets a New folder button beside "Add ... to favorites" (components/ProjectPickerNewFolder.tsx, mounted by components/ProjectPicker.tsx). It opens a name field. Enter or Create makes the folder in the directory on screen and moves the pane into it, so Select makes it the project. Escape closes only the field. On Browse, Tab now moves through the whole picker, from the path field to Add ... to favorites, New folder and the row a caller adds below the panes (the AWS profile choices), instead of closing the picker or dropping into a dialog under it; Enter or Space presses them and Escape still closes the picker (components/ProjectPicker.tsx). A refused name is explained under the field and the typed name stays.

The server side is one owner-only route, POST /api/browse-dirs/create, body {parent, name}, registered next to /api/browse-dirs and living with the listings in file_api/browse.py. It makes exactly one folder level and answers each case the issue names:

Request Answer
Created 201 {path}
Caller is not the dashboard owner 403 owner_only, before the body is read
name empty, ., .., has / or \, would land outside parent (a Windows C:x), or is refused by the filesystem 400 invalid_name
parent missing 400 parent_required
New path in a protected location (for example ~/.aws, or anything inside ~/.ssh) 403 access_denied, nothing created
The parent swapped for a link after the check 409 parent_changed, nothing created
Name already taken 409 already_exists
Parent gone or not a directory, no write permission, anything else 404 not_found, 403 permission_denied, 500 create_failed

The protected-location check is the same sensitive-path floor the listing uses, run on the new path before mkdir. The checked parent is then pinned for the mkdir (a descriptor walked with O_NOFOLLOW on POSIX, a no-rename handle on Windows), the way uploads pin theirs, so a swapped ancestor cannot redirect the folder. The create runs on the path-probe pool like the listing, and each outcome is SEL-audited as browse_create_dir. The API client (api/client/files.ts) gains createBrowseDir, the copy ships in every locale under website/src/i18n, and the dashboard spec and feature map describe the route and the control.

Backwards compatibility

Compatible: a new route and a new control; no existing route, response field or listing changes. One keystroke changes on purpose: on the Browse tab, Tab moves through the picker's controls instead of closing the picker. Escape still closes it. The three tab headers and Select leave the Tab order: they only ever acted on a mouse press, so a keyboard could not press them, and Enter in the path field stays Select's keyboard path.

Removes nothing: the diff only adds a route, a client method, a control and its copy.

Tests

  • test/test_browse_create_dir.py: a create answers 201 with the new path and audits it; it runs off the event loop; an existing name is 409 and left alone; each invalid name (empty, ., .., a slash, a backslash, a NUL, a name too long, which Windows can report as a missing path, a non-string) and a missing parent are 400 with nothing created; on Windows, a drive-qualified name (Z:x) cannot leave the parent; a missing or file parent is 404; a protected new path is 403 before it exists, and the real floor refuses both ~/.aws and a folder inside ~/.ssh; a non-owner is refused before the body is read; a parent without write permission is 403; a parent swapped for a link after the check (a junction on Windows) is 409 with nothing created in the link's target; the route is on the real route table.
  • website/src/test/ProjectPicker.newFolder.test.tsx: create, then the pane is on the new folder ready to Select; each refusal code gets its own copy (the rest take the generic copy), keeps the name and clears on the next edit; Escape closes only the field and returns focus to the path field; a create that answers after the user moved to another folder does not pull the pane back.
  • website/src/test/ProjectPicker.keyboardIsolation.test.tsx, above a real dialog: Tab from the Browse path field reaches the star (Enter favourites the folder) and New folder (Enter opens the field) without closing the picker, and Escape on either closes only the picker; Tab from the name field stays in the field; Escape on the name field, Create or the close control closes only the field; a focused Create stays focusable while its create is out; after the toggle closes the field, Escape dismisses only the picker, not the dialog under it.
  • website/src/test/ProjectPicker.footer.test.tsx: on Browse, Tab from the New folder controls moves through the caller's footer and wraps back to Back.
  • test/test_dashboard_files_composition_contract.py and test/test_json_object_body_guard.py count the new route and record its body cap; website/src/test/ApiClient.refactor.surface.test.ts places createBrowseDir in the client's method order.

Determinism (testing-conventions D12). The new cases read no clock, draw no random number and wait on no timer.

Run Command Result
Python, 20 times at -n0 pytest -n0 test/test_browse_create_dir.py (22 tests, 21 on Linux) 20/20 pass; 5/5 again after both Windows cases stopped skipping
Python, 10 times at -n 4 beside neighbours pytest -n 4 --dist loadgroup test/test_browse_create_dir.py test/test_browse_dirs.py test/test_browse_files.py 10/10 pass
Python, shuffled the 21 node ids in random.Random(seed) order at -n0, seeds 11-15 5/5 pass
Frontend, 20 times serially vitest run src/test/ProjectPicker.newFolder.test.tsx src/test/ProjectPicker.keyboardIsolation.test.tsx --no-file-parallelism (27 tests, before the Tab test was added) 20/20 pass
Frontend, shuffled seeds 1-5, after the keyboard tests vitest run --maxWorkers=1 --sequence.shuffle --sequence.seed=N src/test/ProjectPicker.keyboardIsolation.test.tsx src/test/ImeEnterGuardSites.test.tsx this PR's tests pass on every seed; seed 3 fails one unrelated TagManagerList test, also red on main in that order (#31922)
Backend, 3 times serially pytest -n0 test/test_browse_create_dir.py 3/3 pass
Frontend, 10 times at 4 workers beside neighbours vitest run src/test/ProjectPicker --maxWorkers=4 (11 files, 170 tests) 10/10 pass
Frontend, shuffled the two files above with --sequence.shuffle --sequence.seed=101..105 5/5 pass

Manual verification

Built the SPA from this branch and drove a throwaway kirocrew gateway --test-mode (its own data home and a scratch home folder, listening on loopback only) in headless Chromium: opened the chat composer's project picker, Browse, New folder; an existing name (notes) and .aws in the home folder were each refused with their own copy and the typed name kept; new-app was created and the pane moved into it ready to Select. On disk afterwards new-app existed and .aws was never created. The Windows drive-qualified name case is covered by a Windows-only unit test, which runs on the Windows CI job, not by this run.

Screenshots / video

The name field open under the favourite row.

New folder field open

A name that already exists, explained under the field with the name kept.

Existing name refused

.aws in the home folder, refused as a protected location.

Protected location refused

The pane on the new folder after Create, ready to Select.

Pane on the new folder

Related Issues

Closes #26005

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

@bolichen97
bolichen97 requested review from a team and KiroCrewCaptain as code owners November 6, 2026 18:25
@bolichen97
bolichen97 requested a review from Zedmor November 6, 2026 18:25
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 6, 2026
@bolichen97
bolichen97 enabled auto-merge (squash) November 6, 2026 18:43
@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of a0043e05e1c3591553724b71c008c1ddf6fd0dea — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: the new route is owner-only, checks the sensitive-path floor, and pins its parent for the create. The Tab change is stated as deliberate.

Notes

  • The two removed tests pinned the old behaviour on purpose: Tab closing the picker, and Tab focus staying inside the host dialog. The description calls the Tab change "a maintainer ruling during review", so link that ruling in the PR, because the old pins were decisions and not gaps.

[DESIGN-REVIEWED] a0043e0

Description read

[DESCRIPTION-READ] 2f324b6a3c226e04761de44f4bac36de526298d7319a6cd0342bd2eaf01942eb

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of a0043e05e1c3591553724b71c008c1ddf6fd0dea — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

Nothing blocks. The one part that reaches beyond the Goal, Tab cycling through the whole Browse pane, is declared and can be undone. Its only support is a review ruling the description mentions but does not link.

Notes

  • Item 3 — inherited: its provenance is "Tab on Browse, was a maintainer ruling during review", with no linked comment. It also removes the pin 'closes the picker on a plain Tab (positive control)', which the description does reverse openly ("Tab moves through the picker's controls instead of closing the picker"). Subtraction: if no ruling is on record, narrow useDialogFocusTrap(dropRef, …, { enabled: open && tab === 'browse' … }) so it covers only the open New-folder field and leave the base Tab-closes-picker behaviour alone.

What this change ships

Inventory (7 items) — 6 justified

Intent: let a person make a new folder from the project picker's Browse tab instead of running mkdir in a shell on the gateway host. ADDITION.

  1. A New folder button appears beside "Add … to favorites" on Browse and opens a name field — justified
  2. Create makes one folder in the directory on screen and moves the pane into it, ready to Select — justified
  3. Tab on Browse now moves through the picker's controls and the caller's footer instead of closing the picker — rides along, maintainer ruling not linked
  4. The Browse star becomes a real Tab stop that Enter or Space toggles — rides along
  5. The three tab headers and Select leave the Tab order — rides along
  6. Escape on any Browse control closes only the picker, not a dialog under it — justified
  7. A new owner-only route, POST /api/browse-dirs/create, that checks protected locations, pins the parent before mkdir and records each outcome in the SEL audit log — justified

[FIRST-PRINCIPLES-REVIEWED] a0043e0

Description read

[DESCRIPTION-READ] 79d9d7c03fe201e18effa27b1dc1ac786abb693a9d7d69c5de4b87e213749fff

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ✅ PASS

UX-level review of a0043e05e1c3591553724b71c008c1ddf6fd0dea — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: PASS

Nothing blocks: the blind reader correctly identified New folder, the name field and Create, and said they would use them. The issues below are copy and state notes.

Notes

  • new_folder_protected says "the location is protected", but the blind reader saw notes refused in the same place for another reason and asked "So is this spot 'protected' or only that name?" Name the path instead: "“{{name}}” is a protected folder and can't be created. Choose another name."
  • Create still looks pressable after a refusal (shot-03), because canCreate only checks for a non-empty name. Pressing it again just repeats the same error. Show Create as unavailable (aria-disabled) while create.isError is set and the name hasn't been edited.
  • No screenshot shows the Creating… label. Add one so the pending state can be checked.

[UX-REVIEWED] a0043e0

Description read

[DESCRIPTION-READ] 8ae35f7618422a5aa589e93f66c4a7f8d07d7ccbd7d993d6cbf31a89f1275483

That sha256 names everything this verdict read: the PR description, plus the 7 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed a0043e05e1c3591553724b71c008c1ddf6fd0dea — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] a0043e0

Verdict parsed from the review's SHA-scoped output markers for commit a0043e05e1c3591553724b71c008c1ddf6fd0dea.

False positive or not applicable? A repository writer can comment:
/ai-review override fable a0043e05e1c3591553724b71c008c1ddf6fd0dea false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of a0043e05e1c3591553724b71c008c1ddf6fd0dea and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] a0043e0

False positive or not applicable? A repository writer can comment:
/ai-review override gpt a0043e05e1c3591553724b71c008c1ddf6fd0dea false-positive: <reason, 20 to 500 characters>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 6, 2026
@bolichen97
bolichen97 force-pushed the fix/26005-project-picker-new-folder branch from d25c782 to b7a18c1 Compare November 7, 2026 00:46
@bolichen97

bolichen97 commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor Author
  • Tab pulls focus into the dialog underneath the new-folder form (span=19f07db3c364)

fixed in b7a18c1: the New folder form registers useDialogFocusTrap(formRef, onCancel, {handleEscape: false, restoreFocus: false}), so as the newest armed trap it owns Tab while open; a keyboard-isolation test above a real Modal pins it.
The same head also lets Tab reach the Browse pane's buttons (favourite star, New folder) without closing the picker, with Escape still closing only the picker.

[operator: bolichen97]

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Nov 7, 2026
@bolichen97
bolichen97 force-pushed the fix/26005-project-picker-new-folder branch from b7a18c1 to 4c96efc Compare November 7, 2026 05:20
@bolichen97

Copy link
Copy Markdown
Contributor Author

Rebased: main moved in docs/feature-map/README.md, docs/system-specs/modules/learn-cron-dashboard.md, src/kiro_crew/dashboard/handlers/init.py, website/src/i18n/locales/*.json and website/src/test/ApiClient.refactor.surface.test.ts since this head went green (d18b76e -> 464b547); CI re-running.

[operator: bolichen97]

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: passed Eligible automated validation passed for the current revision readiness: checking Automated validation is still running labels Nov 7, 2026
@bolichen97
bolichen97 force-pushed the fix/26005-project-picker-new-folder branch from 4c96efc to 6b841dd Compare November 7, 2026 09:16
@bolichen97

Copy link
Copy Markdown
Contributor Author

self-added: yes

  • Browse focus trap prevents keyboard navigation through the AWS-profile footer (span=628faec6fd9a)

fixed in 6b841dd: the Browse Tab trap now spans the whole picker (useDialogFocusTrap(dropRef, ...)), so Tab on Browse moves through the caller's footer and wraps back into the pane; ProjectPicker.footer.test.tsx pins it with real Tab presses (New folder Cancel -> footer button -> footer field -> Back).
The New folder form's own trap is removed: the picker-wide trap is the newest one over a host Modal, so Tab still cannot drop into the dialog underneath (the keyboard-isolation tests above a real Modal still pass). The three mousedown-only tab headers leave the Tab order, and the system folder picker button closes the picker on Escape like the pane's buttons.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author

Rebased: main moved in docs/feature-map/README.md, docs/system-specs/modules/learn-cron-dashboard.md, src/kiro_crew/dashboard/handlers/init.py, src/kiro_crew/dashboard/handlers/files.py, test/test_json_object_body_guard.py and website/src/i18n/locales/*.json since this head went green (464b547 -> e30915e); CI re-running.

[operator: bolichen97]

@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Nov 7, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 7, 2026
@bolichen97
bolichen97 force-pushed the fix/26005-project-picker-new-folder branch from 6b841dd to c6419ec Compare November 7, 2026 10:46
@bolichen97

Copy link
Copy Markdown
Contributor Author

self-added: no

  • Windows skips suppress refusal assertions without a missing host capability (span=3bf989d784ae)

fixed in c6419ec: both Windows skips are gone; the parent swap is staged with make_dir_link (a junction on Windows) and the overlong name asserts 400 invalid_name on every OS.
Un-skipping exposed two Windows outcomes, both now mapped: the pin refuses the swapped-in junction with NotADirectoryError, which joins the PinnedPathRefusal branch (a directory behind the name is parent_changed, a file stays not_found); a FileNotFoundError while the parent is still a directory is about the name (Windows reports an over-long path that way), so it answers invalid_name.
The two skips left name real capabilities: POSIX mode bits (and not root) for the read-only parent, and a drive-qualified name, which only exists on Windows.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author

[operator: bolichen97] Backend Tests shard 2 (Linux and Windows) and the Coverage Gate that depends on them fail on test/test_security_gate_liveness.py::test_the_package_stays_within_its_line_budget. That test is red on main too: the security package is 34855 lines on main against a budget of 34845, and this PR does not touch that package. #32183 raises the budget; this PR waits for it to land and then rebases onto it.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Nov 7, 2026
The project picker's Browse tab can list the gateway host's folders and
select one, but not make one, so starting a project in a new folder
meant leaving the dashboard for a shell on the gateway host.

Add a New folder control beside "Add ... to favorites". It opens a name
field; Enter or Create makes the folder in the directory on screen and
lists it, so Select makes it the project. Escape closes only the field,
and a refusal is explained under it with the typed name kept.

The server side is one owner-only route, POST /api/browse-dirs/create,
that makes exactly one folder level. It refuses an existing name, a
name with a slash, a protected location (the sensitive-path floor is
checked on the new path before anything is created) and a parent it
cannot write to, and audits each outcome.

Builds on #26007 by @alparamonov.

Closes #26005
@bolichen97
bolichen97 force-pushed the fix/26005-project-picker-new-folder branch from c6419ec to a0043e0 Compare November 7, 2026 14:34
@bolichen97

Copy link
Copy Markdown
Contributor Author

[operator: bolichen97] Rebased onto d27a5fd, which already raises the security package line budget, so the shard 2 failure in test_the_package_stays_within_its_line_budget (and the Coverage Gate that waits on it) should clear. #32183 was closed without merging; the raise reached main in other PRs. The diff is unchanged. CI is re-running on a0043e0.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Nov 7, 2026

@buluoray buluoray left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 Kiro Crew Auto-Pipeline [operator: buluoray]:

APPROVE

What it does for users: the project picker's Browse tab gets a New folder button. Typing a name and pressing Enter or Create makes one folder in the directory on screen and moves the pane into it, ready to Select. Starting a project in a new folder no longer needs a shell on the gateway host. On Browse, Tab now moves through the picker's own controls instead of closing it, which was a maintainer ruling.

Security review of the new write route (POST /api/browse-dirs/create), checked against source at this head:

  • Owner-only, checked before the body is read: src/kiro_crew/dashboard/file_api/browse.py:238. This is the same gate as /api/file-write, and the listing it sits beside is owner-only too, per the spec's owner-only routes paragraph.
  • The new folder cannot leave the chosen parent. The handler refuses an empty name, ".", "..", "/" and "\" (browse.py:251). The sync body then builds the target with abspath and requires its dirname to equal the realpath'd parent (browse.py:178). That covers a Windows drive-qualified "C:x" and the trailing dot or space Windows strips.
  • The new path cannot reach a sensitive location. The target meets is_sensitive_canonical_path before any mkdir (browse.py:180). The call runs on the probe pool, so it takes the inline pre-resolved gate. That gate case-folds and matches descendants, so a protected parent (anything inside .ssh) is refused as well as a protected leaf (.aws, .netrc).
  • The checked parent cannot be swapped between the check and the create. On POSIX, mkdir runs relative to a descriptor that pin_parent walks with O_NOFOLLOW (browse.py:185). On Windows, a no-delete-share handle is held and its final path is compared with the parent. A platform with neither answers create_failed rather than doing a by-name mkdir (browse.py:198).
  • Every outcome is SEL-audited as browse_create_dir. Nothing existing is loosened: no current route, fence or listing changes.

Evidence:

  • Tests run locally at this head: test_browse_create_dir.py, test_browse_dirs.py, test_dashboard_files_composition_contract.py and test_json_object_body_guard.py gave 150 passed, 2 skipped. The Windows-only cases skip on Linux. The four touched frontend files (ProjectPicker.newFolder, ProjectPicker.keyboardIsolation, ProjectPicker.footer, ApiClient.refactor.surface) gave 221 passed.
  • Mutations, run one at a time with the tree clean after each:
    1. Removing the sensitive-path check reddens test_a_protected_new_path_is_refused_before_it_exists, test_a_folder_inside_a_protected_directory_is_refused and test_a_parent_swapped_for_a_link_after_the_check_creates_nothing.
    2. Replacing the pinned create with a by-name os.mkdir(target) reddens test_a_parent_swapped_for_a_link_after_the_check_creates_nothing.
    3. Removing the owner check reddens test_a_non_owner_is_refused_before_the_body_is_read.
    4. Frontend: dropping the close-on-navigate in browse() (ProjectPicker.tsx:339) reddens "does not pull the pane back when a create answers after the user moved to another folder" and "creates the folder in the directory on screen, then lists it ready to Select".
  • New copy: 11 keys in each of the 13 locales.
  • CI at this head: 83 success, 11 skipped, 0 failing. PR Readiness and both CodeBuild jobs (Linux, Windows) are green. The Design, First Principles, UX, Opus and GPT review lanes all pass. No inline threads.
  • git merge-tree against current main exits 0.

Source-only, not run here: the Windows pin and the drive-qualified-name case. Both have Windows-only tests that the Windows CI job ran green.

Non-blocking follow-ups:

  • The invalid_name copy tells the user to avoid "slashes or special characters". The route only refuses slashes, ".", "..", a NUL and names the filesystem rejects, such as an overlong name. Most special characters are accepted, and the copy does not mention length. Naming what is actually refused would be more accurate.
  • parent_changed has no copy of its own and shows the generic "Couldn't create ... Try again." That is acceptable for a race outcome. It is noted here only so the choice is on record.
  • Like /api/browse-dirs, the create route is not behind the remote-chat guard that /api/file-write uses, so it always acts on the gateway host. That matches the listing the picker shows, so it is not a defect. If the picker ever lists a remote session's filesystem, this route needs the same guard.

@bolichen97
bolichen97 merged commit a2a129f into main Nov 7, 2026
99 checks passed
@bolichen97
bolichen97 deleted the fix/26005-project-picker-new-folder branch November 7, 2026 19:37
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Nov 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Project picker: create a new folder from the Browse tab

2 participants