* fix(backend): make repository setup script failures non-fatal
A failing repository setup script aborted environment preparation and
stopped the agent ("Environment setup failed"), and Resume / Start-fresh
couldn't recover because each relaunch re-ran the same failing script.
runWorktreeSetupScript (run inside worktree creation) treated a non-zero
exit as fatal: it deleted the worktree and returned an error, which
cascaded to EnvPrepareResult{Success: false} and aborted the launch. This
contradicted the already-non-fatal task/profile-level setup script path.
The repo-level script is now non-fatal too: on failure it logs a warning,
records SetupScriptWarning on the worktree, keeps the worktree, and lets
the launch proceed. The env preparer surfaces it on the "Create worktree"
step so the panel shows completed_with_warnings instead of failed. The
full script output still streams to the script_execution chat message.
Adds backend regression tests and an e2e test (verified red before / green
after) asserting the agent still launches when the repo setup script fails.
* fix(backend): preserve base-branch warning when setup script also fails
Addresses review feedback: the post-Create block surfacing SetupScriptWarning
on the "Create worktree" step unconditionally overwrote BaseBranchFallbackWarning
when both occurred, silently dropping the "wrong branch" notice. Append instead
of replace so both survive.
Also updates TestWorktreePreparer_SetupScriptFailure_KeepsCreateWorktreeStepCompleted
to the non-fatal contract: a failed setup script now yields a successful prepare
with a warning on the step, not a failed prepare.
* fix(orchestrator): stop duplicating workflow auto-start prompt on boot-ready drain
When a workflow auto-start prompt failed transiently and was queued by the
retry branches in autoStartStepPrompt, the boot_ready drain later re-inserted
the same user message because executeQueuedMessage always called
CreateUserMessage — producing the two identical Merge prompts observed when
an ACP session resumed after going idle.
Tag the queued message with user_message_recorded when recordAutoStartMessage
has already run before the queue, and have executeQueuedMessage skip its
chat-history insert when the flag is set. The early-busy queue path
(queueAutoStartPromptIfRunning) still queues without the flag so the drain
side keeps owning the record for that path.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix(orchestrator): propagate recordAutoStartMessage result and tighten else-if guard
recordAutoStartMessage now returns bool (true = chat row successfully
inserted). Both queueAutoStartPrompt call sites in autoStartStepPrompt
pass the return value instead of hardcoded true, so user_message_recorded
is only stamped true when CreateUserMessage actually succeeded. A silently
swallowed DB write failure no longer stamps the flag, letting
executeQueuedMessage fall back to inserting the message on drain.
Also scope the debug-log else-if to s.messageCreator != nil so it only
fires when a CreateUserMessage attempt would otherwise have been made.
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>