fix(conpty): stop pinning --effort on hand-driven interactive sessions

Task, ad-hoc, planning, handoff and resume sessions now inherit the CLI
default / the user's /effort setting. List-handler sessions keep their
per-role effort.

docs/explore-notes/conpty-sessions.md

@@ -15,9 +15,9 @@ Autonomous queue tasks are **not** covered here — they stay on the stream-json
 
 | Kind | Hub spec method | Notes |
 |---|---|---|
-| Task session | `GetInteractiveLaunchSpec` | Effort from the task/list model preset |
-| Ad-hoc | `GetAdHocLaunchSpec` | Effort from the global default |
-| Planning | (planning start/resume) | Effort from `PlanningAlias`; uses `--permission-mode default`, **not** `plan` |
+| Task session | `GetInteractiveLaunchSpec` | No `--effort` — CLI default |
+| Ad-hoc | `GetAdHocLaunchSpec` | No `--effort` — CLI default |
+| Planning | (planning start/resume) | No `--effort`; uses `--permission-mode default`, **not** `plan` |
 | List handler | `GetMergeHelperLaunchSpec` / `GetMergeHelperHandoffLaunchSpec` | Model + effort fixed per role (`ModelRegistry.HandlerTriageAlias`/`HandlerWaitAlias`/`HandlerMergeAlias`), **not** from list config; `--permission-mode` via `PermissionModeResolver` (unattended) |
 
 ## ⚠️ Gotcha: never pass task free-text as a CLI argument
@@ -39,9 +39,10 @@ description skips the file **and** the positional arg entirely — it still gets
 
 ## Argument ordering
 
-Every spec passes `--effort <level>` from the relevant model's preset. It **leads** the args —
-except for a fresh task session with a brief, where `--add-dir <sessionDir>` must come first so
-`--effort` (a single-value flag) can sit directly before the positional kickoff.
+Only the list-handler specs pass `--effort` (from the role model's preset). Hand-driven sessions
+(task, ad-hoc, planning, handoff, resume) deliberately pass none, so the CLI default / the user's
+`/effort` setting applies. A positional kickoff must follow a single-value flag (`--session-id`,
+`--append-system-prompt-file`), never a variadic one (`--add-dir`, `--allowedTools`).
 
 `--model` is deliberately **NOT** forced on an interactive session — the user can still switch
 models in the TUI.
@@ -87,7 +88,7 @@ own transcript:
    `TaskRepository.SetInteractiveSessionIdAsync` — **before** the `LaunchSpec` is returned, i.e.
    before the ConPTY host ever spawns `claude`. `BuildFreshTaskArgsAsync` then passes it as
    `--session-id <guid>`, placed as the single-value flag directly before the positional kickoff
-   (or, with no brief, right after `--effort`).
+   (or, with no brief, as the only arg).
 3. **Fresh worktree wins.** `isFreshWorktree` forces `run` to `null` *and* is checked before
    reading `task.InteractiveSessionId`, so a recreated worktree never resumes a stale id from
    either source — it always takes the fresh path, which overwrites the stale
@@ -355,7 +356,7 @@ Autonomous and interactive sessions do **not** share a system prompt. Per start
 |---|---|---|
 | Autonomous run/continue/retry | `TaskRunner.ResolveConfigAsync` → `ClaudeArgsBuilder.Build` | `--append-system-prompt <text>`, recomputed and re-sent on **every** invocation including a `--resume` continue (`PromptKind.System` + improvement/list/task overrides) |
 | Interactive task session, fresh | `InteractiveLaunchSpecService.BuildForTaskAsync` → `BuildFreshTaskArgsAsync` | none — no `--append-system-prompt(-file)` at all |
-| Interactive task session, resume | `InteractiveLaunchSpecService.BuildForTaskAsync` → `WindowsTerminalLauncher.BuildResumeArgs` | none — only `--resume <id>` (+ `--effort`) |
+| Interactive task session, resume | `InteractiveLaunchSpecService.BuildForTaskAsync` → `WindowsTerminalLauncher.BuildResumeArgs` | none — only `--resume <id>` |
 | Ad-hoc directory session | `InteractiveLaunchSpecService.BuildForDirectoryAsync` | none |
 | Planning session start | `InteractiveLaunchSpecService.BuildPlanningStart` → `WindowsTerminalLauncher.BuildPlanningStartArgs` | `--append-system-prompt-file <path>` (`PromptKind.Planning`) |
 | Planning session resume | `InteractiveLaunchSpecService.BuildPlanningResume` → `WindowsTerminalLauncher.BuildPlanningResumeArgs` | none — only `--permission-mode default --allowedTools <planning allowlist> --resume <id>` |

docs/open.md

@@ -32,7 +32,7 @@ Der Sichtprüfungs-Block dieses Batches (Ctrl+K/`#`, Titel-Edit, ConPTY-/Refine-
 Interactive-Chip, Diff-Viewer-Abstände, Manual-Tasks, Vorgaben-pro-Modell-Tabelle) ist von Mika
 verifiziert und deshalb hier entfernt. Es bleiben die Entscheidungen, die daran hängen:
 
-- Interaktive ConPTY-Sessions bekommen `--effort`, aber **kein** `--model` — die Session läuft weiter unter dem Modell aus Mikas Claude-Config, der Effort kommt aus dem Preset des Modells, das ClaudeDo für die Task auflösen würde. Falls ClaudeDo auch interaktiv das Modell erzwingen soll, ist das ein Folge-Task.
+- Interaktive (handgesteuerte) ConPTY-Sessions bekommen weder `--model` noch `--effort` — beides kommt aus Mikas Claude-Config bzw. dem CLI-Default. Nur die List-Handler-Sessions pinnen Modell + Effort pro Rolle.
 - `AppSettings.DefaultMaxTurns` ist jetzt tatsächlich verdrahtet (`ModelPresets.For(..., global.DefaultMaxTurns)` in `TaskRunner.ResolveConfigAsync`): reiner Fallback für ein Modell, das auch nach `ModelRegistry.TryNormalizeAlias` auf keine Preset-Zeile trifft — vorher war das Feld tot (hartkodierte 30). Hat weiterhin keinen eigenen Editor mehr (nur die Preset-Tabelle pro Modell); Spalte könnte später entfallen, falls das nie zutrifft.
 
 ## Beobachtung (offen — Entscheidung Mika)

src/ClaudeDo.Worker/Runner/InteractiveLaunchSpecService.cs

@@ -170,11 +170,6 @@ public sealed class InteractiveLaunchSpecService
         var resolvedClaude = WindowsTerminalLauncher.Resolve(_claudePath)
             ?? throw new InvalidOperationException($"claude executable not found: {_claudePath}");
 
-        // Start the hand-driven session at the effort configured for the model this task would run
-        // under, instead of inheriting whatever the user's global Claude Code config happens to be.
-        // The model itself is deliberately NOT forced here — the user can still switch it in the TUI.
-        var effort = EffortFor(globalSettings, task.Model ?? listConfig?.Model);
-
         // Resume this task's own last interactive conversation when it has one -- it takes
         // precedence over the latest autonomous run's session, since an interactive session is a
         // distinct conversation from an autonomous run even against the same worktree. Fall back
@@ -193,13 +188,13 @@ public sealed class InteractiveLaunchSpecService
         if (resumeSessionId is { Length: > 0 })
         {
             sessionId = resumeSessionId;
-            args = WithEffort(WindowsTerminalLauncher.BuildResumeArgs(resumeSessionId), effort);
+            args = WindowsTerminalLauncher.BuildResumeArgs(resumeSessionId);
         }
         else
         {
             sessionId = Guid.NewGuid().ToString();
             await new TaskRepository(ctx).SetInteractiveSessionIdAsync(taskId, sessionId, ct);
-            args = await BuildFreshTaskArgsAsync(task, effort, sessionId, ct);
+            args = await BuildFreshTaskArgsAsync(task, sessionId, ct);
         }
 
         if (paneId is not null)
@@ -242,8 +237,7 @@ public sealed class InteractiveLaunchSpecService
 
         return new LaunchSpec(
             ctx.WorkingDir, resolvedClaude,
-            WithEffort(WindowsTerminalLauncher.BuildPlanningStartArgs(ctx),
-                       EffortFor(ReadSettings(), ModelRegistry.PlanningAlias)),
+            WindowsTerminalLauncher.BuildPlanningStartArgs(ctx),
             env);
     }
 
@@ -262,8 +256,7 @@ public sealed class InteractiveLaunchSpecService
 
         return new LaunchSpec(
             ctx.WorkingDir, resolvedClaude,
-            WithEffort(WindowsTerminalLauncher.BuildPlanningResumeArgs(ctx.ClaudeSessionId),
-                       EffortFor(ReadSettings(), ModelRegistry.PlanningAlias)),
+            WindowsTerminalLauncher.BuildPlanningResumeArgs(ctx.ClaudeSessionId),
             env);
     }
 
@@ -280,15 +273,13 @@ public sealed class InteractiveLaunchSpecService
             ?? throw new InvalidOperationException($"claude executable not found: {_claudePath}");
 
         await using var ctx = await _dbFactory.CreateDbContextAsync(ct);
-        var settings = await new AppSettingsRepository(ctx).GetAsync(ct);
 
         var env = new Dictionary<string, string>
         {
             ["MCP_TOOL_TIMEOUT"] = McpToolTimeoutMs,
         };
 
-        // No task and no list here — the global default model's preset decides the effort.
-        var args = WithEffort(Array.Empty<string>(), EffortFor(settings, settings.DefaultModel));
+        IReadOnlyList<string> args = [];
         if (paneId is not null)
         {
             // Pre-assigned so the pane entry can carry it -- a handoff records it as the brief's origin.
@@ -322,16 +313,13 @@ public sealed class InteractiveLaunchSpecService
             ?? throw new InvalidOperationException($"claude executable not found: {_claudePath}");
 
         await using var ctx = await _dbFactory.CreateDbContextAsync(ct);
-        var settings = await new AppSettingsRepository(ctx).GetAsync(ct);
         var sessionId = Guid.NewGuid().ToString();
 
-        var model = settings.DefaultModel;
         TaskEntity? task = null;
         if (taskId is not null)
         {
             var taskRepo = new TaskRepository(ctx);
             task = await taskRepo.GetByIdAsync(taskId, ct) ?? throw new KeyNotFoundException($"Task not found: {taskId}");
-            model = task.Model ?? (await new ListRepository(ctx).GetConfigAsync(task.ListId, ct))?.Model;
             await taskRepo.SetInteractiveSessionIdAsync(taskId, sessionId, ct);
         }
 
@@ -352,7 +340,6 @@ public sealed class InteractiveLaunchSpecService
         var args = new List<string>(await PaneSessionArgsAsync(paneId, sessionId, ct))
         {
             "--add-dir", Path.GetDirectoryName(briefPath)!,
-            "--effort", EffortFor(settings, model),
             "--session-id", sessionId,
             $"Read the file {briefPath} first. It is the handoff brief from the session that worked here " +
             "before you. After reading it, continue the work as it describes.",
@@ -379,12 +366,11 @@ public sealed class InteractiveLaunchSpecService
 
         var resolvedClaude = WindowsTerminalLauncher.Resolve(_claudePath)
             ?? throw new InvalidOperationException($"claude executable not found: {_claudePath}");
-        var settings = await new AppSettingsRepository(ctx).GetAsync(ct);
 
         IReadOnlyList<string> args =
         [
             .. await PaneSessionArgsAsync(paneId, sessionId, ct),
-            .. WithEffort(WindowsTerminalLauncher.BuildResumeArgs(sessionId), EffortFor(settings, settings.DefaultModel)),
+            .. WindowsTerminalLauncher.BuildResumeArgs(sessionId),
         ];
         var env = new Dictionary<string, string>
         {
@@ -836,21 +822,6 @@ public sealed class InteractiveLaunchSpecService
     private static string EffortFor(AppSettingsEntity settings, string? model)
         => ModelPresets.For(settings.ModelPresets, model ?? settings.DefaultModel).Effort;
 
-    // Prepends `--effort <level>`. It has to lead: a positional kickoff prompt must stay last, and
-    // it may only follow a single-value flag — a variadic flag would swallow it.
-    private static IReadOnlyList<string> WithEffort(IReadOnlyList<string> args, string effort)
-    {
-        var result = new List<string>(args.Count + 2) { "--effort", effort };
-        result.AddRange(args);
-        return result;
-    }
-
-    private AppSettingsEntity ReadSettings()
-    {
-        using var ctx = _dbFactory.CreateDbContext();
-        return new AppSettingsRepository(ctx).GetAsync().GetAwaiter().GetResult();
-    }
-
     // The task's brief (title + description, unmodified/multi-line) must never travel as a CLI
     // argument: the ConPTY host flattens Args into one command line to spawn the process, and
     // claude re-splits that line on whitespace, so any token in the description that starts with
@@ -858,18 +829,18 @@ public sealed class InteractiveLaunchSpecService
     // the brief at its first line even if it didn't. So it's written to a file instead --
     // same pattern as BuildForMergeHelperAsync / WindowsTerminalLauncher.BuildPlanningStartArgs --
     // and claude is pointed at it with a single-line kickoff. --add-dir exposes the session dir to
-    // Read; --effort (single-value) must sit directly before the positional kickoff so the
+    // Read; --session-id (single-value) must sit directly before the positional kickoff so the
     // preceding variadic --add-dir doesn't swallow the kickoff as another directory.
     // No brief (task has neither a title nor a description) -> no positional arg at all.
     // `--session-id` pre-assigns the claude session id the caller already persisted (see
     // BuildForTaskAsync) so this fresh conversation is resumable from its very first turn --
     // it's a single-value flag, so it may sit directly before the positional kickoff.
     private static async Task<IReadOnlyList<string>> BuildFreshTaskArgsAsync(
-        TaskEntity task, string effort, string sessionId, CancellationToken ct)
+        TaskEntity task, string sessionId, CancellationToken ct)
     {
         var brief = BuildTaskBrief(task);
         if (string.IsNullOrEmpty(brief))
-            return new[] { "--effort", effort, "--session-id", sessionId };
+            return new[] { "--session-id", sessionId };
 
         var sessionDir = Path.Combine(Paths.AppDataRoot(), "task-sessions", task.Id);
         Directory.CreateDirectory(sessionDir);
@@ -879,7 +850,6 @@ public sealed class InteractiveLaunchSpecService
         return new[]
         {
             "--add-dir", sessionDir,
-            "--effort", effort,
             "--session-id", sessionId,
             $"Read the file {briefPath} first. It contains the task you must work on. " +
             "Then, before changing anything, reply with a one-line restatement of what you intend to do " +

tests/ClaudeDo.Worker.Tests/Runner/InteractiveLaunchSpecServiceTests.cs

@@ -215,7 +215,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
         var secondSpec = await svc.BuildForTaskAsync(taskId, CancellationToken.None);
 
         Assert.Equal(_tempDir, secondSpec.Cwd);
-        Assert.Equal(new[] { "--resume", firstSessionId }, ArgsAfterEffort(secondSpec));
+        Assert.Equal(new[] { "--resume", firstSessionId }, HandDrivenArgs(secondSpec));
 
         using var readCtx = _db.CreateContext();
         Assert.Null(await new WorktreeRepository(readCtx).GetByTaskIdAsync(taskId));
@@ -352,7 +352,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
 
         Assert.Equal(_worktreeDir, spec.Cwd);
         Assert.Equal(_claudeStubPath, spec.Exe);
-        Assert.Equal(new[] { "--resume", "sess-123" }, ArgsAfterEffort(spec));
+        Assert.Equal(new[] { "--resume", "sess-123" }, HandDrivenArgs(spec));
         Assert.Equal(InteractiveLaunchSpecService.McpToolTimeoutMs, spec.Env["MCP_TOOL_TIMEOUT"]);
     }
 
@@ -368,7 +368,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
         var svc = BuildService();
         var spec = await svc.BuildForTaskAsync(taskId, CancellationToken.None);
 
-        Assert.Equal(new[] { "--resume", "sess-kept" }, ArgsAfterEffort(spec));
+        Assert.Equal(new[] { "--resume", "sess-kept" }, HandDrivenArgs(spec));
     }
 
     [Fact]
@@ -427,7 +427,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
         var svc = BuildService();
         var spec = await svc.BuildForTaskAsync(taskId, CancellationToken.None);
 
-        Assert.Equal(new[] { "--resume", "interactive-session-1" }, ArgsAfterEffort(spec));
+        Assert.Equal(new[] { "--resume", "interactive-session-1" }, HandDrivenArgs(spec));
     }
 
     [Fact]
@@ -447,7 +447,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
         // start a new conversation.
         var secondSpec = await svc.BuildForTaskAsync(taskId, CancellationToken.None);
 
-        Assert.Equal(new[] { "--resume", firstSessionId }, ArgsAfterEffort(secondSpec));
+        Assert.Equal(new[] { "--resume", firstSessionId }, HandDrivenArgs(secondSpec));
     }
 
     [Fact]
@@ -544,7 +544,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
         var spec = await svc.BuildForTaskAsync(taskId, CancellationToken.None);
 
         // no brief -> no --add-dir, no positional arg -- but a session id is still pre-assigned
-        var rest = ArgsAfterEffort(spec);
+        var rest = HandDrivenArgs(spec);
         Assert.Equal("--session-id", rest[0]);
         Assert.Equal(rest[1], await ReadInteractiveSessionIdAsync(taskId));
     }
@@ -582,7 +582,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
 
         Assert.Equal(_tempDir, spec.Cwd);
         Assert.Equal(_claudeStubPath, spec.Exe);
-        Assert.Empty(ArgsAfterEffort(spec));
+        Assert.Empty(HandDrivenArgs(spec));
         Assert.Equal(InteractiveLaunchSpecService.McpToolTimeoutMs, spec.Env["MCP_TOOL_TIMEOUT"]);
         Assert.Empty(_seeder.Calls);
     }
@@ -761,33 +761,28 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
 
     // ── Merge helper ──
 
-    /// Every ConPTY spec now leads with `--effort <level>` from the per-model preset (the seeded
-    /// settings row has no overrides, so the shipped default for the default model applies).
-    /// Asserts that pair and returns the rest of the args for the test's own comparison.
-    private static string[] ArgsAfterEffort(LaunchSpec spec)
+    /// Hand-driven ConPTY specs never pass `--effort` -- the CLI's own default (or the user's
+    /// /effort setting) applies. Asserts that and returns the args for the test's own comparison.
+    private static string[] HandDrivenArgs(LaunchSpec spec)
     {
-        var args = spec.Args.ToList();
-        Assert.Equal("--effort", args[0]);
-        Assert.Equal(ModelPresets.For(ModelPresets.Defaults, ModelRegistry.DefaultAlias).Effort, args[1]);
-        return args.Skip(2).ToArray();
+        Assert.DoesNotContain("--effort", spec.Args);
+        return spec.Args.ToArray();
     }
 
     /// A fresh (never-resumed) task with a title/description produces
-    /// `--add-dir <sessionDir> --effort <level> --session-id <guid> <kickoff>` -- the brief
+    /// `--add-dir <sessionDir> --session-id <guid> <kickoff>` -- the brief
     /// travels via the file at <sessionDir>/brief.md, never as a raw CLI argument. Asserts that
     /// exact shape and returns (sessionDir, briefPath, kickoff, sessionId) for the test's own
     /// checks. Cleanup happens via `_seededTaskIds` in Dispose (the session dir is keyed by task id).
     private static (string SessionDir, string BriefPath, string Kickoff, string SessionId) ParseFreshTaskArgs(LaunchSpec spec)
     {
         var args = spec.Args.ToList();
-        Assert.Equal(7, args.Count);
+        Assert.Equal(5, args.Count);
         Assert.Equal("--add-dir", args[0]);
         var sessionDir = args[1];
-        Assert.Equal("--effort", args[2]);
-        Assert.Equal(ModelPresets.For(ModelPresets.Defaults, ModelRegistry.DefaultAlias).Effort, args[3]);
-        Assert.Equal("--session-id", args[4]);
-        var sessionId = args[5];
-        var kickoff = args[6];
+        Assert.Equal("--session-id", args[2]);
+        var sessionId = args[3];
+        var kickoff = args[4];
 
         var briefPath = Path.Combine(sessionDir, "brief.md");
         return (sessionDir, briefPath, kickoff, sessionId);
@@ -1562,7 +1557,7 @@ public sealed class InteractiveLaunchSpecServiceTests : IDisposable
 
         var spec = BuildService().BuildPlanningResume(ctx);
 
-        var args = ArgsAfterEffort(spec);
+        var args = HandDrivenArgs(spec);
         Assert.Equal("--permission-mode", args[0]);
         Assert.Equal("default", args[1]);
         Assert.Equal("--allowedTools", args[2]);