preview_merge could only see file-overlap cleanliness via git merge-tree, never
whether the result compiles -- the two costliest findings of the 2026-08-06 batch
run were both merge-tree-clean but build-broken. When a list has a verify command
configured, a clean preview is now additionally materialized (via a commit-tree +
detached scratch worktree, outside the real repo, always cleaned up) and
built/tested there, without ever touching the real working tree.
preview_merge always attempts a verify run when a command is configured;
preview_merge_set only does when its new runVerify parameter is set (default off),
so a set preview never starts N builds unasked. The post-merge verify gate is
unchanged.
Path.GetDirectoryName returned workingDir itself instead of its parent
whenever list.WorkingDir ended in a directory separator, placing
.claudedo-worktrees inside the target repo's own working tree. Add
WorktreeRootResolver to normalize the trailing separator before deriving
the sibling root and to guard that the resulting worktree path never
lands inside workingDir, used by both WorktreeManager and
PlanningSessionManager. Add a startup sweep (LegacyWorktreeFolderRecovery)
that warns when a list's working dir already contains a leftover
.claudedo-worktrees folder from before the fix.
The prime_schedules.prompt_override column existed end-to-end but was
never populated or read. DailyPrepPrompt.BuildPrompt now takes an
optional override and appends it as an extra paragraph after the fixed
prompt (additive, never a replacement, so a user can't disable the
required get_daily_prep_candidates/set_my_day flow). PrimeRunner passes
schedule.PromptOverride through. The Prime tab in Settings now has a
multiline field per schedule wired to a new PromptOverride property on
PrimeScheduleRowViewModel (blank persists as null).
git add -A ran against the shared target checkout, sweeping in
whatever untracked or unrelated modified files a concurrent session
left there and committing them into an unrelated merge. Stage exactly
the paths that were reported conflicted and resolved instead.
A hand-driven task produces no streamed agent output, so the Output tab
opened empty. Bind now selects git for manual/interactive rows and resets
to output for autonomous ones.
DiffEditorSetup is the place for AvaloniaEdit host boilerplate now, so note why
the diff viewer and the merge editor share it but stay separate controls, and
that ModalShell owns the titlebar drag and the OffScreenMargin inset for every
modal.
ModalShell moved its window by assigning Window.Position on every pointer-move,
which bypasses the OS move loop entirely - so Windows never saw a drag gesture
and no snap was ever recognised. Dragging the diff viewer to the top edge did
nothing. The titlebar now calls Window.BeginMoveDrag, which hands the drag to the
window manager and brings back maximise-on-top-edge, half-screen side snap, snap
layouts and the preview overlay. No opt-in flag is needed: the OS only snaps
resizable windows, so the non-resizable modals stay as they are.
The stale comment claiming BeginMoveDrag does not work under Avalonia 12 was
wrong - the method is there in 12.0.4. The VisualRoot cast was the actual problem,
and the window still has to come from TopLevel.GetTopLevel.
Also inset the shell by the window's OffScreenMargin, mirroring what MainWindow
already does for itself. The modals share MainWindow's extended-client-area flags,
so without it a maximised modal overhangs the screen by the invisible resize
border and gets its edges and close button clipped.
Three things were wrong with the fresh side-by-side viewer:
- The panes scrolled independently. The sync hunted the editors' templated
ScrollViewers at load time, but the viewer starts collapsed (no file selected),
so nothing was ever measured, no template existed, and the hook silently
no-opped for the lifetime of the window. Reading now goes through
TextView.ScrollOffsetChanged, which exists from construction; writing goes
through a lazily resolved ScrollViewer.Offset. TextEditor.ScrollToVerticalOffset
is NOT usable here - it is a silent no-op in AvaloniaEdit 12.0.0 even with the
editor templated (verified headlessly).
- Neither pane said what it showed. Added a BASE | WORKTREE header row inside
DiffTextView, sharing the pane grid's columns so each label sits over its editor.
- The layout picker was two lookalike ToggleButtons. It is now a segmented switch
(Border.segmented + Button.segment.active), with wrap demoted to an icon toggle
since it is orthogonal to the layout mode.
The scroll sync, TextMate setup, grammar switching, brush fallback and segment
struct were duplicated between this control and the 3-pane conflict resolver,
which the design doc had used as a copy-paste template. They now live in
DiffEditorSetup; both surfaces migrated. They stay separate controls on purpose -
a read-only two-way renderer over aligned rows with filler lines is not a variant
of a three-way editor over a writable document.
The resolver thereby also picks up the fixed scroll sync and the TryFindResource
brush lookup (its TryGetResource never resolved anything). All four merge tokens
happen to equal their hardcoded fallbacks, so nothing changes on screen.
RunInSlotAsync only logged an unexpected exception, leaving a task stuck
Running in the DB forever with the UI never notified (the raw-SQL queue
claim that put it there never broadcasts). Cancellation is handled
separately and left alone, since the cancel path already wrote the
terminal status.
Every tool description now leads with what the tool does AND when to reach for
it, since MCP clients rank tools by that text. Per-parameter prose moved onto
the parameters as [Description], exhaustive result-shape enumerations and
design/history rationale dropped, and the repeated boilerplate clauses
(lean-task-ref, batch cap, refused-while-Running) pulled into McpToolDocs,
which also documents the style for future tools.
Tool-level description text: 20494 -> 13605 chars (-34%); combined with the new
parameter descriptions 18517 (-10%).
Closes gaps that caused wrong calls rather than just verbose ones:
- list_task_attachments returns metadata only, no file content
- run_task_now shares continue_task's single override slot and throws when busy
- list_runs is ordered oldest-first and feeds get_run
- workingDir on create_list/update_list is an existing local git repo path,
unvalidated until the first task run
- get_task_worktree's behind=0 also means the main ref was unreachable
Removes get_task_status_values: a whole tool entry for static reference text.
GetTask's description is now the canonical place for status meanings.
Audited all 12 prompt kinds against the code they drive. Every real defect sat on
the boundary between prompt text and the --allowedTools the launcher passes.
- Planning: "Use nothing else" after a six-tool list forbade the brainstorming
Skill the same prompt demands two paragraphs earlier. WindowsTerminalLauncher
allowlists mcp__claudedo__*,Read,Grep,Glob,WebFetch,WebSearch,Skill -- name them,
and tell the planner to ground subtasks in the repo with Read/Grep/Glob.
- System: SuggestImprovement is only allowlisted when ParentTaskId is null and
PlanningPhase is None, and TaskRunMcpService throws for any child, but this
prompt reaches every run. Planning children were told to use a tool they lack.
- MergeHelperExecute: derived "effective max-turns" from task/list/preset by hand,
which misses TaskRunner's MaxTurnsCeiling clamp. Call get_effective_run_config
instead -- built for exactly this and reports the clamp.
- MergeHelperExecute: quoted the override-slot error as the raw lowercase throw
rather than the string ExternalMcpService actually surfaces.
- Refine: listed Read/Grep/Glob unconditionally though RefinePrompt.BuildArgs only
appends them when a repo is available.
Two findings deliberately left open, both needing a code decision rather than a
prompt edit: the System prompt's worktree claim is false for a list without a
WorkingDir (task runs in a plain sandbox dir), and 'fable' is missing from both the
prompt's cost ordering and ModelRegistry.ByCostAscending.
The execute prompt told the handler to wait with timeoutSeconds up to 170 -- a
leftover from the retired MCP_TOOL_TIMEOUT=200000ms era. The real server-side
clamp is TaskWaitMcpTools.MaxTimeoutSeconds = 900 and every launcher sets
930000ms, so the handler was making ~5x the wait_for_task_change calls it needed
and burning turns on re-waiting.
It also never passed treatWaitingForChildrenAsBusy, and only waited on ids that
were Queued or Running. A task with children reports "changed" the moment it
reaches WaitingForChildren, so such a task both dropped out of the wait set and
signalled completion early -- the handler could reach review/merge while
children were still running.
RenderTemplate replaced tokens one key at a time over a StringBuilder, so a
token appearing inside an already-substituted value got substituted again on a
later pass. The prompt briefs only escaped this because their callers happen to
pass "tasks" last -- reordering the dictionary or adding a fourth token would
have started rewriting user-authored task descriptions, which after the enhance
phase carry file paths and config snippets.
Single-pass regex over the template instead; unknown tokens still pass through.
mika kuns