Follow up to #57704
This makes sure that we offer a worktree creation option in case
resolving the default branch fails
Release Notes:
- git: Fixed an issue where worktree creation would not be possible if
resolving default branch fails
Adds opt-in rendering for Markdown frontmatter metadata blocks in
Markdown Preview and agent markdown.
- Simple `key: value` metadata blocks now render as a two-column table,
while more complex metadata falls back to a code-style block.
- Metadata block content and key/value rows are parsed in the parser
step, and the request layout simply takes over rendering.
<img width="1288" height="436" alt="image"
src="https://github.com/user-attachments/assets/b35b949a-8bc4-47db-82ef-ed835e9ac06f"
/>
Release Notes:
- Added support for rendering Markdown frontmatter metadata blocks in
Markdown Preview and Agent Panel.
Provide a way to prevent GPUI from creating AccessKit adapters, and
enable this in Zed.
This will allow us to test AccessKit support in Zed without rolling it
out more broadly, while we gain confidence in the implementation in
GPUI.
I've also added a log statement
## Motivation (i.e. a mini post-mortem about the #56065 panics)
Merging #56065 caused some nasty panics in nightly. This was caused by a
bug in the logic for selecting a focus node for a `TreeUpdate`.
AccessKit panics when an invalid `TreeUpdate` is provided.
My assumption was that, since Zed uses no a11y APIs, and also that
essentially 0 zed users would have AT apps running, that merging this PR
would have no effect on the behaviour of Zed itself. However, two issues
combined to cause the panics:
- It seems like many people (everyone?) on mac gets the activation
callback called by accesskit_macos. A quick search suggests this might
be due to password managers searching for password fields, but not sure
how true that is.
- The bug in question related to *forgetting to check* whether a node
used a11y APIs, so we *were* pushing non-empty `TreeUpdate`s
As a (probably temporary) defensive measure, I added a function to try
to detect the bad cases and fix them. But it would be lovely if this
could live in AccessKit itself, since it would mean we wouldn't have to
do the check twice (once in GPUI, once in AccessKit). This would also
help prevent drift when updating accesskit versions if new invariants
are added.
We also cannot protect against this with `catch_unwind`, since we use
`panic=abort`. So our only option unfortunately is to temporarily
disable AccessKit until we know our implementation is stable.
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
---------
Co-authored-by: zed-zippy[bot] <234243425+zed-zippy[bot]@users.noreply.github.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#56357
Release Notes:
- Fixed bug where devcontainers were not respecting override_command
Selecting text inside the hover documentation or git popups did not
scroll the popup when the drag passed the visible area, so any text
below the area could not be selected with the mouse.
The popup's container already had a `ScrollHandle` wired to its
scrollbar and wheel scrolling, but the inner `MarkdownElement` was
constructed without one. That left it in the default
`AutoscrollBehavior::Propagate` mode, which routes drag-autoscroll
requests to the editor-wide autoscroll listener, which is a listener
that does not exist inside a floating popover, so the requests were
silently dropped.
Passing the popup's existing `ScrollHandle` into the `MarkdownElement`
switches it to `AutoscrollBehavior::Controlled`, which scrolls the
popup's own container directly during a drag. The markdown preview view
already uses this same pattern.
Release Notes:
- Fixed hover documentation and git popups not scrolling while selecting
text with the mouse
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
I ran into this while using lazygit: the Shift navigation keys were
being captured by Zed to scroll the terminal buffer instead of being
passed through to the program.
Capturing them makes sense in the normal terminal case, where the user
wants to scroll and Shift has no other meaning. But in alternate-screen
TUIs like lazygit, less, or neovim, terminal scrollback isn't relevant,
so we can forward these keys to the program while it's open.
Release Notes:
- Fixed Shift+Up, Shift+Down, Shift+Home, and Shift+End in terminal TUIs
like lazygit, less, and neovim.
---------
Co-authored-by: John Tur <john-tur@outlook.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
Some best-effort tracking of an allowed-list of agents (to avoid
grabbing sensitive data) just to get basic data on general usage
patterns.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
---------
Co-authored-by: morgankrey <morgan@zed.dev>
Closes#56493
In Helix select mode, pressing `a` after a selection (e.g. `v a`) placed
the cursor one column too far to the right.
Bound `a` to `vim::HelixAppend` in the `helix_select` keymap so it
matches the behavior in `helix_normal`, and added a regression test for
the `v a` case.
Release Notes:
- Fixed cursor placement after pressing `a` in Helix select mode.
---------
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
We were missing a check when setting the focused node ID, causing a
panic in accesskit's internal validation logic. Fixes the bug, and adds
some defensive logic to help mitigate potential future issues.
You can reproduce the panic by:
- turning on a screenreader
- launching zed
- opening a menu and pressing tab
This change fixes it.
Release Notes:
- N/A or Added/Fixed/Improved ...
Stacked on top of #57430.
When the `sandboxing` feature flag is on (macOS only), agent-run
terminal commands are launched under `/usr/bin/sandbox-exec` with a
per-command Seatbelt policy:
- **Reads:** any path on the filesystem.
- **Writes:** each project worktree plus a per-command `$TMPDIR`.
- **Network:** blocked.
The model can request relaxations on individual calls via three new
input flags on the `terminal` tool:
- `allow_network: true`
- `allow_fs_write: true`
- `unsandboxed: true`
Each one triggers a separate always-prompt user approval (bypassing any
`always_allow` rules, since escalation is a stronger trust boundary than
the baseline command approval). The flags are silently ignored when
sandboxing is off, so the model can't surreptitiously change runtime
behavior by setting them in the no-op case.
The per-command tempdir is provisioned regardless of sandbox state so
the model can't infer whether the sandbox is in effect by looking at
`$TMPDIR`.
The Seatbelt config file's lifetime is tied to the `Terminal` entity via
an opaque RAII handle (`SandboxConfigHandle = Box<dyn Any + Send>`), so
cancellation paths drop it automatically.
### Notes for review
- The writable scope passed to the sandbox is the project's worktree
paths plus the per-command tempdir, **not** the resolved `cd` working
directory — `cd` is model-controlled, and using it as the writable scope
would let the model widen its own write permissions outside the project.
- The new tool flags are always present in the JSON schema (with
`#[serde(default)]`), even when the sandbox prompt section isn't
rendered. The system prompt only documents them when the section is
present, so the model shouldn't try to use them when it's not — but the
schema doesn't actively reject them, just ignores them.
- I opted not to add an end-to-end test for the escalation prompt path:
the existing `FakeThreadEnvironment` ignores the new params and toggling
the feature flag in tests is more wiring than felt worth it for a first
cut. The pure-function pieces (`sandbox_approval_title`, schema,
deserialization) are covered, and the sandbox crate itself has
end-to-end tests that actually invoke `sandbox-exec`.
Release Notes:
- N/A
---------
Co-authored-by: MartinYe1234 <52641447+MartinYe1234@users.noreply.github.com>
Stacked on top of #57429.
Adds a new `sandboxing` feature flag (off for staff by default) and a
single source-of-truth helper `sandboxing_enabled(cx) = cfg!(target_os =
"macos") && cx.has_flag::<SandboxingFeatureFlag>()`. When the helper
returns true, the agent's system prompt gains a new `## Terminal
sandbox` section that:
- Lists each worktree's absolute path as a writable directory.
- Describes the per-command `$TMPDIR` scratch directory.
- States that outbound network access is blocked.
- Documents the three per-command flags (`allow_network`,
`allow_fs_write`, `unsandboxed`) the model can request to relax the
sandbox.
- Tells the model the section is stable for the duration of the
conversation.
When the flag is off, the section is omitted entirely — no mention of
sandboxing at all.
No behavior change to terminal execution yet; that's the next PR in the
stack.
Three new tests cover: section omitted when `sandboxing: false`, section
rendered with all worktrees + flag docs when `sandboxing: true`, and the
zero-worktrees case.
Release Notes:
- N/A
---------
Co-authored-by: MartinYe1234 <52641447+MartinYe1234@users.noreply.github.com>
Co-authored-by: Martin Ye <martin@zed.dev>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] ~Unsafe blocks (if any) have justifying comments~ (N/A)
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior (didn't add new test for
parsing SHA-256 - not sure if would be desired)
- [x] Performance impact has been considered and is acceptable
Closes#24070
Upgrades git2 from 0.20.1 to 0.21.0 with the `unstable-sha256` feature -
adds ability to open and work with git repositories using the SHA-256
object format. `Oid::from_str` now detects 64-char hex strings to parse
SHA-256 OIDs correctly.
Also adapts to breaking API changes in 0.21.0:
`Remote::url()` and `Commit::message()` both now return `Result`.
Release Notes:
- Added support for opening SHA-256 object format git repositories
Adds a new `sandbox` crate that wraps shell invocations under macOS's
`sandbox-exec(1)` with a Seatbelt policy built from per-command
permissions:
- Reads are unrestricted.
- Writes are restricted to a caller-provided list of directories (plus
the standard `/dev/*` write targets).
- Network access and unrestricted filesystem writes must be opted into
per command.
`wrap_invocation(program, args, writable_dirs, permissions)` returns the
new program/args plus a `SeatbeltConfigFile` RAII handle that deletes
the on-disk policy file when dropped — callers hold it for the lifetime
of the spawned command.
No callers yet — this is the first of three stacked PRs. The second
wires the sandbox state into the agent's system prompt behind a feature
flag; the third wires the actual wrapping into the agent terminal tool.
The macOS-only dependencies (`tempfile`, `anyhow`) are gated by
`target.'cfg(target_os = "macos")'` so the crate is empty on other
platforms.
Includes 14 tests covering both the generated Seatbelt policy text and
end-to-end behavior (actually invoking `sandbox-exec` and asserting
reads/writes succeed or fail per policy).
Release Notes:
- N/A
Removes a leftover test label added by #56065
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
Follow-up to #57772.
While comparing the `edit_file_tool_streaming/medium_insertions`
benchmark in Criterion and xctrace, I noticed the DP scoring loop was
spending most of its time on inlined `Matrix::get`/`Matrix::set` access
in `StreamingDiff::push_new`.
This changes the loop to split the previous and current score columns
once per DP column and then index those column slices directly. This
keeps the scoring algorithm the same, but hoists the column offset
calculation out of the inner loop and gives the compiler clearer
aliasing information between the immutable previous column and mutable
current column.
In xctrace, the hottest DP scoring line dropped from 10,407 samples
before to 9,409 samples after on
`edit_file_tool_streaming/medium_insertions`, and total
`StreamingDiff::push_new` leaf samples dropped from 10,800 to 9,822.
This is consistently faster in the end-to-end `edit_file_tool`
benchmark, and gives a small speed-up for the targeted workload without
changing the scoring algorithm.
### Results
Benchmarked with `release-fast` against a clean baseline using
Criterion’s `before_slice` baseline.
#### `streaming_diff_push_new`
| Fixture | Before median | After median | Change |
| --- | ---: | ---: | ---: |
| `tiny_function_rewrite` | `7.6317 ms` | `6.4951 ms` | 10.8% faster |
| `small_function_rewrite` | `45.961 ms` | `44.839 ms` | No significant
change |
| `medium_many_small_changes` | `99.694 ms` | `104.83 ms` | 5.2% slower
|
| `medium_insertions` | `95.909 ms` | `83.093 ms` | 13.5% faster |
#### `edit_file_tool_streaming`
| Fixture | Before median | After median | Change |
| --- | ---: | ---: | ---: |
| `tiny_function_rewrite` | `1.1110 ms` | `1.0617 ms` | 7.1% faster |
| `small_function_rewrite` | `2.4554 ms` | `2.2751 ms` | 7.1% faster |
| `medium_many_small_changes` | `87.272 ms` | `77.065 ms` | 11.5% faster
|
| `medium_insertions` | `92.469 ms` | `87.478 ms` | 5.9% faster |
The lower-level `streaming_diff_push_new` benchmark is mixed, but the
end-to-end `edit_file_tool_streaming` workload improves across all
fixtures.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A or Added/Fixed/Improved ...
Two model updates:
- **Free**: Big Pickle now has a max output tokens of 32k as per
49991c8f8f
- **Go**: added Qwen3.7 Max as per
4f1d5c511a
and
1554a5a82e (diff-a7ee7cf35b40335095a44cfaac16041313c36322cd14544ddad0289ce69131ab).
Tested this successfully by running a simple "_rename this variable for
me. add a function. delete the function_" test
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- OpenCode: updated models (added Qwen3.7 Max, updated Big Pickle token
counts)
GPUI AccessKit integration
This PR is replacing #51097 , and is much more limited in scope. This PR
*ONLY* adds AccessKit support to GPUI, and doesn't touch Zed. Once this
lands, we can start adding aria attributes to Zed's components.
This PR is the first step to addressing #41138 .
Release Notes:
- N/A or Added/Fixed/Improved ...
---------
Co-authored-by: John Tur <john-tur@outlook.com>
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Co-authored-by: Agus Zubiaga <agus@zed.dev>
This PR is a follow-up to #54206 to make it so we only respect the
feature flag overrides for Zed staff (or when running a development
build).
Previously just the UI portion of the feature flags was disabled for
non-staff, but manipulating the settings file by hand would still allow
setting overrides that would be respected.
The only other setting on the "Developer" page of the settings UI was
the performance profiler. This was previously only available for staff,
as the entire "Developer" page was limited to staff, but it seems
reasonable to allow non-staff to see this.
Release Notes:
- N/A
This PR uses the status toast for the Skill Creator confirming action as
opposed the regular message notification. Aside from it looking a bit
better, it's also auto-dismissed, which is preferred in this case.
Release Notes:
- Improved skill creation toast confirmation by making it
auto-dismissed.
---------
Co-authored-by: Martin Ye <martin@zed.dev>
Don't panic when a diff transform boundary falls inside a multi-byte
UTF-8 character. Round split points up to the next char boundary and
advance past any transform boundaries skipped by that adjustment.
Closes FR-16
Release Notes:
- Fixed a crash when deleting words near inline diff boundaries
containing multi-byte characters
## Motivation
Miniprofs of Zed dropping frames showed foreground thread stalls of
711.56 ms, 641.01 ms, and 221.99 ms from the `edit_file_tool`. The new
benchmark isolated `StreamingDiff::push_new` as one of the expensive
phases so this PR aims to speed it up so we can avoid dropping more
frames in the future.
## Benchmark methodology
I had an agent create a Criterion suite with four deterministic Rust
edit fixtures: tiny localized rewrite, small localized rewrite, many
small changes, and helper block insertions.
Benchmarked with:
```sh
cargo bench -p streaming_diff --bench streaming_diff --profile release-fast -- --warm-up-time 1 --measurement-time 2
```
The benchmark binary was also recorded under `xctrace` CPU Counters to
inspect CPU samples before and after the change.
## Results
`StreamingDiff::push_new` improved across all fixtures:
| Benchmark | Before | After | Improvement |
|---|---:|---:|---:|
| `tiny_function_rewrite` | ~10.81 ms | ~6.91 ms | ~36% faster |
| `small_function_rewrite` | ~51.02 ms | ~35.39 ms | ~31% faster |
| `medium_many_small_changes` | ~130.71 ms | ~83.83 ms | ~36% faster |
| `medium_insertions` | ~120.52 ms | ~79.90 ms | ~34% faster |
The `xctrace` baseline showed samples in `Hash::hash`,
`RandomState::hash_one`, `HashMap::insert`, and
`RawTable::reserve_rehash`. After replacing the map with two `Vec<u32>`
row buffers, the hash table frames disappeared from the top samples.
The speedup comes from removing hash work from the DP inner loop and
replacing scattered hash table probes with contiguous row buffer access
(better cache hits 🏎️). The trace did not include direct cache miss
counts, but this layout is likely more cache friendly because it
replaces hash table access with a continuous line of memory (the
`Vec`s).
Release Notes:
- N/A
## Summary
- Track existing skill file contents and instruction bodies before
migrating Rules to Skills
- Skip writing migrated skills when an equivalent skill already exists
- Cover duplicate detection across differently named skills
## Tests
- cargo test -p prompt_store rules_to_skills_migration
Release Notes:
- Fixed duplicate Skills being created when migrating Rules.
Until now, the cloud-hosted model list was only refreshed in response to
events that exercise the LLM token (a `UserUpdated` push, an
organization change, or `PrivateUserInfoUpdated`). If a user wasn't
actively using AI features around the time we shipped new models, the
list could stay stale until they restarted Zed.
This is the second step toward fixing that, after #57078 made the cloud
websocket reconnect on its own. We now treat each successful (re)connect
as a hint that the server state may have changed, so possibly new model
definitions will be available, and trigger a model list refresh.
The trigger is a new `Client::cloud_connection_id()` watch that bumps a
counter each time the websocket handshake completes.
`CloudLanguageModelProvider::State` subscribes to it and, on every tick
after the initial `0`, schedules a debounced refresh (with jitter, so we
don't have all active clients trying to reconnect at the same time after
we deploy in cloud).
Closes CLO-713.
Release Notes:
- The list of Zed hosted models is now refreshed automatically, without
requiring a restart
<img width="620" height="172" alt="Screenshot 2026-05-27 at 12 08 26 PM"
src="https://github.com/user-attachments/assets/226b3d0c-003b-44ac-a16f-10af4f2952b3"
/>
Add command palette actions for opening global and project-specific
AGENTS.md files
Closes AI-324
Release Notes:
- Added commands to open global and project-specific AGENTS.md rules
Stop trying to add new watches for 5 seconds after receiving the "OS
file watch limit reached" error.
This was flooding the logs and many pointless syscalls.
Related to #57422, #57042, FR-18
Release Notes:
- Improved file watcher behavior when the OS file watch limit is
reached.
Closes#47251
Fix dot (`.`) repeat not correctly repeating the last change after
replaying a macro (`@register`)
([#47251](https://github.com/zed-industries/zed/issues/47251))
When replaying a macro that contains text insertions,
`replay_insert_event` calls `handle_input` directly and never emits
`InputHandled`, so the `observe_insertion` subscription never fires.
This left the dot register stale — `.` after `@register` would repeat an
earlier change instead of the last one made by the macro.
Fix by calling `observe_insertion` explicitly in the
`ReplayableAction::Insertion` branch of `Replayer::next`.
Release Notes:
- Fixed dot (`.`) repeat not repeating the last change made by a macro
(`@register`).
---------
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
Tiny UI tweak here to improve the label/traffic lights alignment in the
skills creator header.
Release Notes:
- N/A
Co-authored-by: Martin Ye <martin@zed.dev>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Needed for use in cloud
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Release Notes:
- agent: Fixed an issue where commit message generation would not
respect instructions from global `AGENTS.md`
---------
Co-authored-by: Richard Feldman <oss@rtfeldman.com>
Also adds a menu item called "Rules Library" that takes to the
documentation, which will explain the removal of the feature as a whole.
Release Notes:
- N/A
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Extends the same `~/.agents/skills` special case that
`create_directory`, `edit_file`, and `write_file` already use to the
path tools, so the agent can copy or move skills into and out of the
global skills folder and delete individual skills or skill resources
beneath it.
- `copy_path` now allows source and/or destination to be a descendant of
`~/.agents/skills`, going through `fs::copy_recursive` directly when one
side is outside the project.
- `move_path` now allows source and/or destination to be a descendant of
`~/.agents/skills`, going through `fs.rename` directly when one side is
outside the project. Moving the `~/.agents/skills` root itself is
rejected.
- `delete_path` now allows deleting any file or directory beneath
`~/.agents/skills`, going through `fs.remove_dir` / `fs.remove_file`
directly. Deleting the `~/.agents/skills` root itself is rejected.
- All three tools still always prompt for approval on agent-skill paths,
even when default tool permissions are set to allow.
- Added shared helpers in `tool_permissions.rs` for resolving global
skill descendants and rejecting operations on the skills root where
needed.
- Added tests covering copying and moving skills in both directions,
deleting a global skill directory/file, and rejecting deletion of the
skills root.
Release Notes:
- Agent can now copy or move skills into and out of `~/.agents/skills`
and delete individual skills, with an explicit confirmation prompt for
each operation
Follow-up to #57466 simplifying `load_skill_frontmatter` rather than
fixing a separate bug.
#57466 fixed the chunked-read loop so it truncates accumulated bytes at
the frontmatter boundary, preventing `str::from_utf8` from failing on a
multi-byte grapheme split across chunks. That fix is correct, but the
chunked read it's working around isn't pulling its weight:
`MAX_SKILL_FILE_SIZE` is 100KB and there's already a metadata pre-check,
so "stop reading early once we've seen the closing `---`" saves at most
~25 pages per file while forcing the use of `open_sync`, a hand-rolled
loop, and a long comment about `smol::unblock` vs the GPUI test
scheduler.
This PR:
- replaces the chunked `open_sync` + read loop with a single
`fs.load(...)` call (the same primitive `read_skill_body` already uses);
- deletes `closing_delimiter_end`, `SKILL_READ_CHUNK_SIZE`, the
`std::io::{self, Read}` import, and the `Parking forbidden` paragraph;
- tightens the metadata pre-check so a metadata error (permissions, I/O,
etc.) bails out instead of silently falling through to a blind read;
- removes the now-obsolete
`test_load_skill_frontmatter_with_emoji_at_chunk_boundary` test — by
construction, a single full read can't split a UTF-8 sequence at any
boundary.
Closes AI-303
Release Notes:
- N/A
The Skill Creator window previously couldn't be resized — its minimum
size was set to the same dimensions as its initial size, so users
couldn't make it smaller or shorter. Its contents also didn't scroll, so
on smaller windows the Skill Content editor would be squeezed below its
usable size.
This change:
- Lowers the window's minimum size to 500×420 so it can be resized in
both directions.
- Wraps the form body in a scrollable container with a vertical
scrollbar.
- Replaces `flex_1 + min_h_0` on the form fields and Skill Content
wrapper with `flex_grow + flex_shrink_0`. Combined with the existing
`min_h(160)` on the body editor, this means:
- When there's extra vertical space, the Skill Content editor grows to
fill it.
- When the window is short, the form keeps its content at natural size
and the outer container scrolls instead of crushing the editor.
Closes AI-315
Release Notes:
- Made the Skill Creator window resizable and its contents scrollable
when the window is smaller than the form.
It look like this:
<img width="1698" height="688" alt="grafik"
src="https://github.com/user-attachments/assets/02a37271-63d3-42da-887f-e17b31e8d9ca"
/>
The idea is to avoid people turning on fast mode without understanding
the financial implications. It also clarifiers (in the BYOK case) why
they might not see a difference between fast mode enabled and disabled.
Release Notes:
- N/A
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Co-authored-by: zed-zippy[bot] <234243425+zed-zippy[bot]@users.noreply.github.com>
async_tasks can build a cycle with the scheduler where the runnable
itself keeps a reference to the scheduler and the scheduler keeping
references to the runnable. This effectively can cause scheduler / task
leaks, most notably in test environments, so we break this cycle by
using weak pointers instead.
Release Notes:
- N/A or Added/Fixed/Improved ...
Same mechanism as for BYOK: `service_tier == priority`. Most of the work
is already done. When validating this in manual testing, I noticed we
get back `service_tier == auto` in the response, unlike in the regular
OpenAI API scenario with BYOK, but apparently [it doesn't mean priority
tier wasn't
applied](https://github.com/openai/codex/issues/14204#issuecomment-4033184620).
It's not a hard confirmation, but the model does seem to respond faster
when I toggle fast mode on.
Release Notes:
- Added Fast Mode (priority service tier) support to OpenAI models used
through the ChatGPT subscription provider.
Maps the existing `Speed::Fast` plumbing to OpenAI's `service_tier:
"priority"`, which matches what "fast mode" in Codex does. Relevant docs
[here](https://platform.openai.com/docs/api-reference/chat/create#chat-create-service_tier).
Like for the existing Anthropic fast mode we have a
`Model::supports_priority` method for the variants on
https://openai.com/api-priority-processing. Pro, nano, and legacy gpt-4
are excluded; Custom defaults to false.
This is gated to staff only for now (not in this diff, but the existing
fast mode feature), until we have the mechanism to require confirmation
before you enable fast mode.
Release Notes:
- Added support for Fast Mode (priority service tier) on the OpenAI API
provider.
> [!NOTE]
> Authored by Claude, but fixes a real issue I encountered; I have
reviewed the diff
Some npm registries return `npm info --json` with non-string values in
the `time` map. JFrog Artifactory emits `"unpublished": null`, and npm
itself represents `unpublished` as an object for packages that have had
versions unpublished.
Because `NpmInfo.time` was typed `HashMap<String, String>`, serde
aborted the entire deserialization with `invalid type: null, expected a
string`. As a result `select_npm_package_version` never ran, and
language servers installed via npm (for example the tsgo extension,
`tailwindcss-language-server`, and `json-language-server`) failed to
start with an error like:
```
Failed to start language server "tsgo": invalid type: null, expected a string at line 100 column 23
```
Only version keys in `time` are ever read (to honor npm's `before`
cutoff), so this deserializes the map leniently: keep the string-valued
entries and drop the rest. The field type and all downstream logic are
unchanged. Added a regression test covering a `time` map containing
`"unpublished": null`.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the UI/UX checklist
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed language servers failing to install when the npm registry
returns non-string values (such as `"unpublished": null`) in package
`time` metadata
