feat(maestro): force kill deletes dirty worktrees

Previously --force left the worktree on disk so the operator could
inspect it. Now it removes the worktree unconditionally — the
confirmation prompt already shows the breakdown, so the decision
happens before the delete, not after.

Natural exit still leaves dirty worktrees on disk (no operator
confirmation in that path).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

crates/maestro-cli/src/cli.rs

@@ -133,7 +133,7 @@ pub struct KillArgs {
     /// Session ID to kill.
     pub session_id: Uuid,
     /// End the session even when its worktree holds unsaved work (#81). The
-    /// worktree is left on disk — never removed.
+    /// worktree is removed along with the session.
     #[arg(long)]
     pub force: bool,
 }

crates/maestro-daemon/src/session.rs

@@ -1545,15 +1545,10 @@ impl SessionManager {
     /// uncommitted tracked changes, untracked non-ignored files, or commits
     /// reachable from no remote-tracking ref. Nothing is killed and nothing
     /// is removed; the error names what blocked it and how much of it there
-    /// is. The session stays as the operator's handle on their own unfinished
-    /// work, so ending a session that still matters becomes a deliberate
-    /// `--force`, which leaves the tree alone rather than destroying it
-    /// (WORKTREES.md).
+    /// is.
     ///
-    /// With `force` (#81) the refusal — and nothing else — is overridden:
-    /// the session ends and the guarded worktree is left on disk, never
-    /// removed. A tree whose guard is clear never needed the override; force
-    /// removes it as usual.
+    /// With `force` (#81) the refusal is overridden: the session ends and
+    /// the worktree is removed, even when it holds unsaved work.
     ///
     /// This is also how an already-dead session is dismissed. tmux is only
     /// asked to kill something it still has — calling `kill-session` on a
@@ -1588,7 +1583,7 @@ impl SessionManager {
             sessions.remove(&id)
         };
         let record = removed.ok_or(SessionError::NotFound(id))?;
-        cleanup_record(&self.inner, id, record, alive, guarded).await;
+        cleanup_record(&self.inner, id, record, alive).await;
         Ok(())
     }
 
@@ -2972,8 +2967,8 @@ async fn mark_session_dead_with_worktree_settlement(
 
 /// Apply the exit rule (#82) to a vanished session's worktree: probe what
 /// would be lost through the kill's own predicate, then remove a pristine
-/// tree. Anything holding unsaved work is left exactly where it is — the
-/// same thing a forced kill does with a guarded tree (#81).
+/// tree. Anything holding unsaved work is left on disk — there is no
+/// operator confirmation on the natural-exit path.
 ///
 /// Returns whether the worktree was fully accounted for. When it wasn't,
 /// nothing was destroyed and the caller keeps the reference on the record.
@@ -3028,16 +3023,11 @@ async fn settle_worktree_on_exit(
 /// gone from the authoritative in-memory map, so a stray FIFO or sidecar can't
 /// resurrect it. `kill_tmux` is `true` for an explicit kill and `false` for the
 /// liveness reaper (tmux already lost the session).
-///
-/// `keep_worktree` is `true` for a forced kill (#81) over a guarded tree: the
-/// session ends but the worktree is left on disk, so only the removal is
-/// skipped.
 async fn cleanup_record(
     inner: &Arc<Inner>,
     id: Uuid,
     record: SessionRecord,
     kill_tmux: bool,
-    keep_worktree: bool,
 ) {
     if kill_tmux {
         let tmux = inner.tmux.clone();
@@ -3059,17 +3049,7 @@ async fn cleanup_record(
         tracing::warn!(error = ?e, %id, "cleanup: failed to delete sidecar");
     }
     if let (Some(base), Some(path)) = (&record.worktree_base_repo, &record.worktree_path) {
-        if keep_worktree {
-            tracing::info!(
-                %id,
-                path = %path.display(),
-                "forced kill: worktree left on disk"
-            );
-        } else if let Err(e) = inner.worktree_mgr.remove(base, path) {
-            // The kill already passed the unsaved-work guard (#79), so this
-            // removal is unconditional; the manager's own nested-worktree
-            // refusal is all that can still leave a tree on disk, and it
-            // says where.
+        if let Err(e) = inner.worktree_mgr.remove(base, path) {
             tracing::warn!(
                 %id,
                 path = %path.display(),

docs/WORKTREES.md

@@ -75,27 +75,17 @@ The point of refusing rather than cleaning up is that the session stays as the
 operator's handle on their own unfinished work. A session that is killed while
 its worktree still matters leaves a directory nobody is looking at.
 
-A consequence, accepted deliberately: a feature branch almost always has unpushed
-commits, so `--force` is the ordinary way to end a session, not an exception.
-That is only safe because force never removes what blocked.
-
 ### `--force`
 
-`--force` overrides the refusal and nothing else. It ends the session and
-**leaves the worktree on disk**, untouched. Force never removes a tree that
-would have blocked. A tree whose guard is clear never needed the override, so a
-forced kill removes it as usual — force changes what happens to blocked work,
-not ordinary cleanup.
+`--force` overrides the refusal and **removes the worktree**, even when it holds
+unsaved work. The operator sees the breakdown in the confirmation prompt and
+decides; once confirmed, the tree is gone.
 
 In the web client the kill button sends the unforced `DELETE` straight away —
 there is no pre-confirmation, because the daemon is the one that decides. Only
 the refusal raises a prompt, and that prompt names what is in the way and offers
 "Force kill", which repeats the request with `force=true`.
 
-Ending a session and destroying its work are two separate acts. Maestro only
-performs the first one. Removing a tree it left behind is the operator's own
-`git worktree remove`, chosen while looking at exactly what would be lost.
-
 ### Exiting on its own
 
 A session that ends by itself — the command finished, the operator typed `exit` —
@@ -113,17 +103,12 @@ once. A removed tree drops off the record, so dismissing the session afterwards
 does not refuse: there is no guard left to trip.
 
 A tree that was left alone stays attached to the dead record, where dismissing it
-later goes through `--force`, which leaves it on disk too. The same holds when
-the probe itself fails: nothing is destroyed.
+later goes through `--force`, which removes it. The same holds when the probe
+itself fails: nothing is destroyed.
 
-## Nothing is removed on the operator's behalf
+## Trees left behind by natural exit
 
-There is no store of worktrees the daemon is holding, no list endpoint, and no
-reclaim command. A tree left behind by a forced kill or a dirty exit is an
-ordinary git worktree on disk; `git worktree list` finds it and
-`git worktree remove` removes it.
-
-That is deliberate. A tracked list of trees pending deletion is a second place
-the truth can live, and it goes stale the moment somebody removes a directory by
-hand. Git already knows which worktrees exist, and removing one is not a
-decision Maestro is better placed to make.
+A forced kill always removes the worktree. A natural exit with unsaved work
+leaves the tree on disk — there is no operator confirmation in that path. Those
+trees are ordinary git worktrees; `git worktree list` finds them and
+`git worktree remove` removes them.