* 🎛️ perf: Narrow Composer Subscriptions to Streaming State
* ✅ test: Cover Composer Subscription Refactor's Behavioral Changes
Add regression tests for the previously-uncovered changed behavior in the
composer-subscriptions refactor:
- useLatestMessageMeta: exact projected field set, null on empty cache, and
referential stability + no re-render across token-only cache writes.
- useGetLatestMessage: call-time tail read, stable callback identity with no
re-render on cache writes, Recoil-snapshot sibling-branch resolution, null
with no conversation.
- useSubmitMessage: reads the tail at call time and appends it to root when
missing (and does not when present or absent) — the reconcile branch the
prior test skipped via an early return.
- useHandleKeyUp: ArrowUp in an empty composer clicks the latest message's edit
control, with the null / missing-control / non-empty-composer guards.
- useAskAnswerMode (new spec): liveAsk is projected through the
findLiveAskUserQuestion select, null when empty/disabled.
* 🎨 style: Fix import order in useLatestMessage spec
- Updated the Skills component to include a check for `skillsEnabled` from agent capabilities, ensuring skills are only displayed when both permissions and capabilities allow.
- Modified the `useHandleKeyUp` hook to prevent triggering skill commands when skills are disabled, enhancing user experience and preventing errors.
- Added tests to verify that skills commands do not trigger when skills are disabled, ensuring robust functionality.
- Refactored the `useSideNavLinks` hook to incorporate skills capability checks, ensuring consistent access control across the application.
* feat: add $ command popover for manual skill invocation
Adds a new `$` command trigger in the chat textarea that opens a
searchable popover listing user-invocable skills. Selecting a skill
inserts `$skill-name` into the message and enables the skills badge
on the ephemeral agent. Follows the same patterns as `@` mentions
and `/` prompts.
* test: update useHandleKeyUp tests for $ skill command
Add showSkillsPopoverFamily, dollarCommand, and SKILLS permission
to the store/recoil/access mocks. Include test cases for the $
trigger, toggle gating, and permission gating.
* fix: address review findings in SkillsCommand
- Exclude auto-mode skills from popover (isUserInvocable now
returns false for InvocationMode.auto)
- Rewrite JSDoc to match the corrected filter logic
- Guard ArrowUp/ArrowDown against NaN when matches is empty
- Replace useRecoilState with useSetRecoilState for ephemeralAgent
to avoid subscribing to unrelated agent state changes
- Move skills guard into setter callback and drop ephemeralAgent
from handleSelect dependency array
- Add e.preventDefault() for Tab key alongside Enter
- Render error state (com_ui_skills_load_error) on query failure
- Render empty state (com_ui_skills_empty) when no matches
- Hoist ScrollText icon to module-level constant
- Export isUserInvocable for testability
* fix: address follow-up review findings
- Differentiate empty-catalog ("No skills yet") from no-match
search ("No skills found") using existing com_ui_no_skills_found
- Clamp activeIndex when matches shrink from search filtering to
prevent silent Enter/Tab failures
- Add early return after Escape handler to skip redundant checks
- Reorder package imports shortest-to-longest after react
- Add fast-typing test case for $ command ("$sk" at position 3)
* fix: gate $ command on assistants endpoint, fix Tab on empty matches
- Block $ popover on assistants/azureAssistants endpoints where
ephemeralAgent is not sent in the submission payload
- Allow Tab to close the popover when matches is empty instead of
trapping keyboard focus
- Add endpoint gating tests for $ on both assistants endpoints
* fix: reset skills popover on assistants switch + paginate skills query
- Close $ skills popover when endpoint switches to assistants or
azureAssistants, mirroring the existing + command reset
- Switch SkillsCommand from a single-page list query to the
cursor-paginated useSkillsInfiniteQuery and auto-fetch all pages
so client-side search covers the full catalog instead of only
the first 100 entries
- Show the spinner during background page fetches and suppress the
empty-state copy until paging completes
- Add test for popover reset on endpoint switch
* fix: prevent currency hijack and form submit on $ command
- Reject the $ trigger when fast-typed text after $ does not start
with a lowercase letter, so currency input like $100 or $5.99
no longer opens the skills popover and clears the textarea
- preventDefault() on Enter when the popover has no matches so
the surrounding chat form does not submit when the user dismisses
the popover via Enter
- Add tests for $100 and $5.99 currency inputs
* fix: defer $ popover to second keystroke and stop pagination on errors
- Defer opening the skills popover until a letter follows the $
character. Bare $ no longer triggers the popover, so starting a
message with $100, $5.99, or $EUR is fully preserved (the
textarea is not cleared by useInitPopoverInput on the first
keystroke). $a, $skill, $my-skill still open as expected.
- Add a sticky paginationBlockedRef circuit breaker on the auto
fetchNextPage effect so a transient page request error cannot
spin into an unbounded retry loop when isError flips back to
false on the next attempt.
- Update tests: bare $ no longer triggers, $a does, currency cases
remain blocked.
* feat: scaffold structured manual-skill channel for follow-up PR
Add a per-conversation pendingManualSkillsByConvoId atom family
that SkillsCommand appends to on selection. This is the writer
half of the structured channel that will let a follow-up PR
deterministically prime SKILL.md as a meta user message before
the LLM turn (mirroring Claude Code's `/skill` invocation), so
the backend never has to regex-parse `$name` out of user text.
The submit pipeline does not yet read this atom; the textual
`$skill-name ` insertion remains the authoritative signal until
the follow-up wires the read + reset on submit. Reset is already
wired into useClearStates so the atom does not leak across
conversation switches.
* test: cover SkillsCommand selection-flow contract
Add a component test that locks in the contract the follow-up
manualSkills PR has to honor when a user picks a skill in the $
popover:
- Pushes the skill name onto pendingManualSkillsByConvoId (the
per-conversation structured channel), with dedup
- Flips ephemeralAgent.skills to true via the callback-form setter
- Inserts $skill-name into the textarea as cosmetic confirmation
- Closes the popover via setShowSkillsPopover(false)
- Renders nothing when the popover atom is false
Mocks recoil setters and the skills query so the test exercises
the real component logic without spinning up the full provider
stack.
* revert: drop $ defer-until-letter guard for sibling consistency
Bring $ in line with @, /, and +: open the popover on the bare
trigger character. The earlier defer guard kept currency input
like \$100 from clobbering the textarea, but it broke the natural
"type \$ to browse skills" UX and was inconsistent with the other
trigger chars, none of which gate on the follow-up character
(/path/to/file, @username, +1 all clear the textarea the same way).
The currency hijack remains recoverable via Escape.
Drop the corresponding paste-protection tests for \$100 and \$5.99,
restore the bare-\$ trigger test.
* refactor: Improve UX for Command Popovers
* Added loading state handling in Mention and PromptsCommand components to display a spinner when data is being fetched.
* Refactored onFocus logic to clear the textarea and set the search value based on command character input.
* Introduced a new `isLoading` state in the useMentions hook to manage loading indicators across multiple data queries.
* Added unit tests for the useHandleKeyUp hook to ensure command triggering works correctly under various conditions.
* ci: useHandleKeyUp tests for command navigation
* Added tests to ensure that the command popovers do not trigger when the cursor is mid-text after pressing ArrowLeft or Delete.
* Updated the shouldTriggerCommand function to refine the conditions under which commands are triggered based on cursor position.
* Improved agent query handling in useMentions hook for better performance and clarity.
* refactor: Optimize Mention and PromptsCommand Components
* Refactored Mention and PromptsCommand components to utilize Recoil state for popover visibility, improving state management and reducing prop drilling.
* Simplified onFocus logic to enhance user experience when interacting with command inputs.
* Added unit tests for useHandleKeyUp to ensure proper command handling and popover visibility based on user input.
* Improved performance by memoizing popover state and reducing unnecessary re-renders.
* fix: Address review findings for command popover refactor
- Fix endpointType regression: add effectiveEndpointByIndex selector
that returns endpointType ?? endpoint, matching the original ChatForm
guard for custom endpoints proxying assistants
- Extract duplicated initInputRef callback into shared useInitPopoverInput
hook, used by both Mention and PromptsCommand
- Add navigation keys (ArrowLeft, ArrowRight, ArrowDown, Home, End,
Delete) to invalidKeys to prevent false popover triggers
- Add endpoint gating tests for assistants/azureAssistants blocking the
+ command
- Remove unused _index param from MentionContent