@os-eco/overstory-cli 0.8.0 → 0.8.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (44) hide show
  1. package/README.md +3 -1
  2. package/agents/builder.md +2 -2
  3. package/agents/lead.md +2 -2
  4. package/agents/merger.md +2 -2
  5. package/agents/orchestrator.md +1 -1
  6. package/agents/reviewer.md +2 -2
  7. package/agents/scout.md +2 -2
  8. package/agents/supervisor.md +3 -3
  9. package/package.json +1 -1
  10. package/src/agents/overlay.test.ts +42 -0
  11. package/src/agents/overlay.ts +1 -0
  12. package/src/commands/dashboard.test.ts +86 -0
  13. package/src/commands/dashboard.ts +8 -4
  14. package/src/commands/feed.test.ts +8 -0
  15. package/src/commands/inspect.test.ts +156 -1
  16. package/src/commands/inspect.ts +19 -4
  17. package/src/commands/replay.test.ts +8 -0
  18. package/src/commands/sling.test.ts +34 -10
  19. package/src/commands/sling.ts +249 -136
  20. package/src/commands/status.test.ts +77 -0
  21. package/src/commands/status.ts +6 -3
  22. package/src/commands/stop.test.ts +186 -4
  23. package/src/commands/stop.ts +46 -14
  24. package/src/commands/trace.test.ts +8 -0
  25. package/src/config.test.ts +63 -0
  26. package/src/config.ts +29 -5
  27. package/src/index.ts +2 -2
  28. package/src/logging/theme.ts +4 -0
  29. package/src/runtimes/codex.test.ts +22 -8
  30. package/src/runtimes/codex.ts +21 -16
  31. package/src/runtimes/connections.test.ts +74 -0
  32. package/src/runtimes/connections.ts +34 -0
  33. package/src/runtimes/registry.test.ts +1 -1
  34. package/src/runtimes/registry.ts +2 -0
  35. package/src/runtimes/sapling.test.ts +1237 -0
  36. package/src/runtimes/sapling.ts +698 -0
  37. package/src/runtimes/types.ts +45 -0
  38. package/src/types.ts +7 -1
  39. package/src/watchdog/daemon.ts +34 -0
  40. package/src/watchdog/health.test.ts +102 -0
  41. package/src/watchdog/health.ts +140 -69
  42. package/src/worktree/process.test.ts +101 -0
  43. package/src/worktree/process.ts +111 -0
  44. package/src/worktree/tmux.ts +5 -0
package/README.md CHANGED
@@ -19,6 +19,7 @@ Requires [Bun](https://bun.sh) v1.0+, git, and tmux. At least one supported agen
19
19
  - [GitHub Copilot](https://github.com/features/copilot) (`copilot` CLI)
20
20
  - [Codex](https://github.com/openai/codex) (`codex` CLI)
21
21
  - [Gemini CLI](https://github.com/google-gemini/gemini-cli) (`gemini` CLI)
22
+ - [Sapling](https://github.com/nichochar/sapling) (`sp` CLI)
22
23
 
23
24
  ```bash
24
25
  bun install -g @os-eco/overstory-cli
@@ -181,6 +182,7 @@ Overstory is runtime-agnostic. The `AgentRuntime` interface (`src/runtimes/types
181
182
  | Copilot | `copilot` | (none — `--allow-all-tools`) | Active development |
182
183
  | Codex | `codex` | OS-level sandbox (Seatbelt/Landlock) | Active development |
183
184
  | Gemini | `gemini` | `--sandbox` flag | Active development |
185
+ | Sapling | `sp` | `.sapling/guards.json` | Active development |
184
186
 
185
187
  ## How It Works
186
188
 
@@ -280,7 +282,7 @@ overstory/
280
282
  metrics/ SQLite metrics + pricing + transcript parsing
281
283
  doctor/ Health check modules (11 checks)
282
284
  insights/ Session insight analyzer for auto-expertise
283
- runtimes/ AgentRuntime abstraction (registry + adapters: Claude, Pi, Copilot, Codex, Gemini)
285
+ runtimes/ AgentRuntime abstraction (registry + adapters: Claude, Pi, Copilot, Codex, Gemini, Sapling)
284
286
  tracker/ Pluggable task tracker (beads + seeds backends)
285
287
  mulch/ mulch client (programmatic API + CLI wrapper)
286
288
  e2e/ End-to-end lifecycle tests
package/agents/builder.md CHANGED
@@ -20,7 +20,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
20
20
 
21
21
  ## overlay
22
22
 
23
- Your task-specific context (task ID, file scope, spec path, branch name, parent agent) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `ov sling` and tells you WHAT to work on. This file tells you HOW to work.
23
+ Your task-specific context (task ID, file scope, spec path, branch name, parent agent) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `ov sling` and tells you WHAT to work on. This file tells you HOW to work.
24
24
 
25
25
  ## constraints
26
26
 
@@ -108,7 +108,7 @@ You are an implementation specialist. Given a spec and a set of files you own, y
108
108
 
109
109
  ## workflow
110
110
 
111
- 1. **Read your overlay** at `.claude/CLAUDE.md` in your worktree. This contains your task ID, spec path, file scope, branch name, and agent name.
111
+ 1. **Read your overlay** at `{{INSTRUCTION_PATH}}` in your worktree. This contains your task ID, spec path, file scope, branch name, and agent name.
112
112
  2. **Read the task spec** at the path specified in your overlay. Understand what needs to be built.
113
113
  3. **Load expertise** via `ml prime [domain]` for domains listed in your overlay. Apply existing patterns and conventions.
114
114
  4. **Implement the changes:**
package/agents/lead.md CHANGED
@@ -43,7 +43,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
43
43
 
44
44
  ## overlay
45
45
 
46
- Your task-specific context (task ID, spec path, hierarchy depth, agent name, whether you can spawn) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `ov sling` and tells you WHAT to coordinate. This file tells you HOW to coordinate.
46
+ Your task-specific context (task ID, spec path, hierarchy depth, agent name, whether you can spawn) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `ov sling` and tells you WHAT to coordinate. This file tells you HOW to coordinate.
47
47
 
48
48
  ## constraints
49
49
 
@@ -160,7 +160,7 @@ Action: Full Scout → Build → Verify pipeline. Spawn scouts for exploration,
160
160
 
161
161
  Delegate exploration to scouts so you can focus on decomposition and planning.
162
162
 
163
- 1. **Read your overlay** at `.claude/CLAUDE.md` in your worktree. This contains your task ID, hierarchy depth, and agent name.
163
+ 1. **Read your overlay** at `{{INSTRUCTION_PATH}}` in your worktree. This contains your task ID, hierarchy depth, and agent name.
164
164
  2. **Load expertise** via `ml prime [domain]` for relevant domains.
165
165
  3. **Search mulch for relevant context** before decomposing. Run `ml search <task keywords>` and review failure patterns, conventions, and decisions. Factor these insights into your specs.
166
166
  4. **Load file-specific expertise** if files are known. Use `ml prime --files <file1,file2,...>` to get file-scoped context. Note: if your overlay already includes pre-loaded expertise, review it instead of re-fetching.
package/agents/merger.md CHANGED
@@ -19,7 +19,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
19
19
 
20
20
  ## overlay
21
21
 
22
- Your task-specific context (task ID, branches to merge, target branch, merge order, parent agent) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `overstory sling` and tells you WHAT to merge. This file tells you HOW to merge.
22
+ Your task-specific context (task ID, branches to merge, target branch, merge order, parent agent) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `overstory sling` and tells you WHAT to merge. This file tells you HOW to merge.
23
23
 
24
24
  ## constraints
25
25
 
@@ -97,7 +97,7 @@ You are a branch integration specialist. When workers complete their tasks on se
97
97
 
98
98
  ## workflow
99
99
 
100
- 1. **Read your overlay** at `.claude/CLAUDE.md` in your worktree. This contains your task ID, the branches to merge, the target branch, and your agent name.
100
+ 1. **Read your overlay** at `{{INSTRUCTION_PATH}}` in your worktree. This contains your task ID, the branches to merge, the target branch, and your agent name.
101
101
  2. **Read the task spec** at the path specified in your overlay. Understand which branches need merging and in what order.
102
102
  3. **Review the branches** before merging:
103
103
  - `git log <target>..<branch>` to see what each branch contains.
@@ -31,7 +31,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
31
31
 
32
32
  ## overlay
33
33
 
34
- Your task-specific context (task ID, file scope, spec path, branch name, parent agent) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `ov sling` and tells you WHAT to work on. This file tells you HOW to work.
34
+ Your task-specific context (task ID, file scope, spec path, branch name, parent agent) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `ov sling` and tells you WHAT to work on. This file tells you HOW to work.
35
35
 
36
36
  ## constraints
37
37
 
@@ -16,7 +16,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
16
16
 
17
17
  ## overlay
18
18
 
19
- Your task-specific context (task ID, code to review, branch name, parent agent) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `overstory sling` and tells you WHAT to review. This file tells you HOW to review.
19
+ Your task-specific context (task ID, code to review, branch name, parent agent) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `overstory sling` and tells you WHAT to review. This file tells you HOW to review.
20
20
 
21
21
  ## constraints
22
22
 
@@ -95,7 +95,7 @@ You are a validation specialist. Given code to review, you check it for correctn
95
95
 
96
96
  ## workflow
97
97
 
98
- 1. **Read your overlay** at `.claude/CLAUDE.md` in your worktree. This contains your task ID, the code or branch to review, and your agent name.
98
+ 1. **Read your overlay** at `{{INSTRUCTION_PATH}}` in your worktree. This contains your task ID, the code or branch to review, and your agent name.
99
99
  2. **Read the task spec** at the path specified in your overlay. Understand what was supposed to be built.
100
100
  3. **Load expertise** via `ml prime [domain]` to understand project conventions and standards.
101
101
  4. **Review the code changes:**
package/agents/scout.md CHANGED
@@ -16,7 +16,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
16
16
 
17
17
  ## overlay
18
18
 
19
- Your task-specific context (what to explore, who spawned you, your agent name) is in `.claude/CLAUDE.md` in your worktree. That file is generated by `overstory sling` and tells you WHAT to work on. This file tells you HOW to work.
19
+ Your task-specific context (what to explore, who spawned you, your agent name) is in `{{INSTRUCTION_PATH}}` in your worktree. That file is generated by `overstory sling` and tells you WHAT to work on. This file tells you HOW to work.
20
20
 
21
21
  ## constraints
22
22
 
@@ -97,7 +97,7 @@ You perform reconnaissance. Given a research question, exploration target, or an
97
97
 
98
98
  ## workflow
99
99
 
100
- 1. **Read your overlay** at `.claude/CLAUDE.md` in your worktree. This contains your task assignment, spec path, and agent name.
100
+ 1. **Read your overlay** at `{{INSTRUCTION_PATH}}` in your worktree. This contains your task assignment, spec path, and agent name.
101
101
  2. **Read the task spec** at the path specified in your overlay.
102
102
  3. **Load relevant expertise** via `ml prime [domain]` for domains listed in your overlay.
103
103
  4. **Explore systematically:**
@@ -31,7 +31,7 @@ These are named failures. If you catch yourself doing any of these, stop and cor
31
31
 
32
32
  ## overlay
33
33
 
34
- Unlike the coordinator (which has no overlay), you receive your task-specific context via the overlay CLAUDE.md at `.claude/CLAUDE.md` in your worktree root. This file is generated by `ov supervisor start` (or `ov sling` with `--capability supervisor`) and provides:
34
+ Unlike the coordinator (which has no overlay), you receive your task-specific context via the overlay CLAUDE.md at `{{INSTRUCTION_PATH}}` in your worktree root. This file is generated by `ov supervisor start` (or `ov sling` with `--capability supervisor`) and provides:
35
35
 
36
36
  - **Agent Name** (`$OVERSTORY_AGENT_NAME`) -- your mail address
37
37
  - **Task ID** -- the issue you are assigned to
@@ -163,7 +163,7 @@ Before spawning, check `ov status` to ensure non-overlapping file scope across a
163
163
 
164
164
  ## workflow
165
165
 
166
- 1. **Receive the dispatch.** Your overlay (`.claude/CLAUDE.md`) contains your task ID and spec path. The coordinator sends you a `dispatch` mail with task details.
166
+ 1. **Receive the dispatch.** Your overlay (`{{INSTRUCTION_PATH}}`) contains your task ID and spec path. The coordinator sends you a `dispatch` mail with task details.
167
167
  2. **Read your task spec** at the path specified in your overlay. Understand the full scope of work assigned to you.
168
168
  3. **Load expertise** via `ml prime [domain]` for each relevant domain. Check `{{TRACKER_CLI}} show <task-id>` for task details and dependencies.
169
169
  4. **Analyze scope and decompose.** Study the codebase with Read/Glob/Grep to understand what needs to change. Determine:
@@ -418,7 +418,7 @@ You are long-lived within a project. You survive across batches and can recover
418
418
  - **Checkpoints** are saved to `.overstory/agents/$OVERSTORY_AGENT_NAME/checkpoint.json` before compaction or handoff. The checkpoint contains: agent name, assigned task ID, active worker IDs, task group ID, session ID, progress summary, and files modified.
419
419
  - **On recovery**, reload context by:
420
420
  1. Reading your checkpoint: `.overstory/agents/$OVERSTORY_AGENT_NAME/checkpoint.json`
421
- 2. Reading your overlay: `.claude/CLAUDE.md` (task ID, spec path, depth, parent)
421
+ 2. Reading your overlay: `{{INSTRUCTION_PATH}}` (task ID, spec path, depth, parent)
422
422
  3. Checking active group: `ov group status <group-id>`
423
423
  4. Checking worker states: `ov status`
424
424
  5. Checking unread mail: `ov mail check --agent $OVERSTORY_AGENT_NAME`
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@os-eco/overstory-cli",
3
- "version": "0.8.0",
3
+ "version": "0.8.3",
4
4
  "description": "Multi-agent orchestration for AI coding agents — spawn workers in git worktrees via tmux, coordinate through SQLite mail, merge with tiered conflict resolution. Pluggable runtime adapters for Claude Code, Pi, and more.",
5
5
  "author": "Jaymin West",
6
6
  "license": "MIT",
@@ -875,6 +875,48 @@ describe("formatQualityGatesCapabilities", () => {
875
875
  });
876
876
  });
877
877
 
878
+ describe("INSTRUCTION_PATH placeholder", () => {
879
+ test("defaults to .claude/CLAUDE.md when instructionPath is not set", async () => {
880
+ const config = makeConfig({
881
+ baseDefinition: "Read your overlay at {{INSTRUCTION_PATH}} in your worktree.",
882
+ });
883
+ const output = await generateOverlay(config);
884
+
885
+ expect(output).toContain("Read your overlay at .claude/CLAUDE.md in your worktree.");
886
+ expect(output).not.toContain("{{INSTRUCTION_PATH}}");
887
+ });
888
+
889
+ test("uses custom instructionPath when set", async () => {
890
+ const config = makeConfig({
891
+ instructionPath: "SAPLING.md",
892
+ baseDefinition: "Read your overlay at {{INSTRUCTION_PATH}} in your worktree.",
893
+ });
894
+ const output = await generateOverlay(config);
895
+
896
+ expect(output).toContain("Read your overlay at SAPLING.md in your worktree.");
897
+ expect(output).not.toContain("{{INSTRUCTION_PATH}}");
898
+ expect(output).not.toContain(".claude/CLAUDE.md");
899
+ });
900
+
901
+ test("INSTRUCTION_PATH in base definition replaced throughout (multiple occurrences)", async () => {
902
+ const config = makeConfig({
903
+ instructionPath: "AGENTS.md",
904
+ baseDefinition: "Step 1: read {{INSTRUCTION_PATH}}.\nContext is in {{INSTRUCTION_PATH}}.",
905
+ });
906
+ const output = await generateOverlay(config);
907
+
908
+ expect(output).not.toContain("{{INSTRUCTION_PATH}}");
909
+ expect(output.split("AGENTS.md").length - 1).toBeGreaterThanOrEqual(2);
910
+ });
911
+
912
+ test("no unreplaced INSTRUCTION_PATH placeholders in final output", async () => {
913
+ const config = makeConfig({ instructionPath: "SAPLING.md" });
914
+ const output = await generateOverlay(config);
915
+
916
+ expect(output).not.toContain("{{INSTRUCTION_PATH}}");
917
+ });
918
+ });
919
+
878
920
  describe("quality gate placeholders in base definitions", () => {
879
921
  test("QUALITY_GATE_INLINE in base definition gets replaced", async () => {
880
922
  const config = makeConfig({
@@ -320,6 +320,7 @@ export async function generateOverlay(config: OverlayConfig): Promise<string> {
320
320
  "{{QUALITY_GATE_CAPABILITIES}}": formatQualityGatesCapabilities(config.qualityGates),
321
321
  "{{TRACKER_CLI}}": config.trackerCli ?? "sd",
322
322
  "{{TRACKER_NAME}}": config.trackerName ?? "seeds",
323
+ "{{INSTRUCTION_PATH}}": config.instructionPath ?? ".claude/CLAUDE.md",
323
324
  };
324
325
 
325
326
  let result = template;
@@ -409,6 +409,92 @@ describe("renderAgentPanel", () => {
409
409
  // dimBox.vertical is a dimmed ANSI string — present in output
410
410
  expect(out).toContain(dimBox.vertical);
411
411
  });
412
+
413
+ test("renders Live column header (not Tmux)", () => {
414
+ const data = makeDashboardData({});
415
+ const out = renderAgentPanel(data, 100, 12, 3);
416
+ expect(out).toContain("Live");
417
+ expect(out).not.toContain("Tmux");
418
+ });
419
+
420
+ test("shows green dot for headless agent with alive PID", () => {
421
+ const alivePid = process.pid; // own PID — guaranteed alive
422
+ const data = {
423
+ ...makeDashboardData({}),
424
+ status: {
425
+ currentRunId: null,
426
+ agents: [
427
+ {
428
+ id: "sess-h1",
429
+ agentName: "headless-worker",
430
+ capability: "builder",
431
+ worktreePath: "/tmp/wt/headless",
432
+ branchName: "overstory/headless/task-1",
433
+ taskId: "task-h1",
434
+ tmuxSession: "", // headless
435
+ state: "working" as const,
436
+ pid: alivePid,
437
+ parentAgent: null,
438
+ depth: 0,
439
+ runId: null,
440
+ startedAt: new Date(Date.now() - 10_000).toISOString(),
441
+ lastActivity: new Date().toISOString(),
442
+ escalationLevel: 0,
443
+ stalledSince: null,
444
+ transcriptPath: null,
445
+ },
446
+ ],
447
+ worktrees: [],
448
+ tmuxSessions: [], // no tmux sessions
449
+ unreadMailCount: 0,
450
+ mergeQueueCount: 0,
451
+ recentMetricsCount: 0,
452
+ },
453
+ };
454
+ const out = renderAgentPanel(data, 100, 12, 3);
455
+ // Green ">" for alive headless agent
456
+ expect(out).toContain(">");
457
+ expect(out).toContain("headless-worker");
458
+ });
459
+
460
+ test("shows red dot for headless agent with dead PID", () => {
461
+ const deadPid = 2_147_483_647;
462
+ const data = {
463
+ ...makeDashboardData({}),
464
+ status: {
465
+ currentRunId: null,
466
+ agents: [
467
+ {
468
+ id: "sess-h2",
469
+ agentName: "dead-headless", // short enough to not be truncated
470
+ capability: "builder",
471
+ worktreePath: "/tmp/wt/dead-headless",
472
+ branchName: "overstory/dead-headless/task-2",
473
+ taskId: "task-h2",
474
+ tmuxSession: "", // headless
475
+ state: "working" as const,
476
+ pid: deadPid,
477
+ parentAgent: null,
478
+ depth: 0,
479
+ runId: null,
480
+ startedAt: new Date(Date.now() - 10_000).toISOString(),
481
+ lastActivity: new Date().toISOString(),
482
+ escalationLevel: 0,
483
+ stalledSince: null,
484
+ transcriptPath: null,
485
+ },
486
+ ],
487
+ worktrees: [],
488
+ tmuxSessions: [],
489
+ unreadMailCount: 0,
490
+ mergeQueueCount: 0,
491
+ recentMetricsCount: 0,
492
+ },
493
+ };
494
+ const out = renderAgentPanel(data, 100, 12, 3);
495
+ expect(out).toContain("x");
496
+ expect(out).toContain("dead-headless");
497
+ });
412
498
  });
413
499
 
414
500
  describe("openDashboardStores", () => {
@@ -42,6 +42,7 @@ import { createTrackerClient, resolveBackend } from "../tracker/factory.ts";
42
42
  import type { TrackerIssue } from "../tracker/types.ts";
43
43
  import type { EventStore, MailMessage, StoredEvent } from "../types.ts";
44
44
  import { evaluateHealth } from "../watchdog/health.ts";
45
+ import { isProcessAlive } from "../worktree/tmux.ts";
45
46
  import { getCachedTmuxSessions, getCachedWorktrees, type StatusData } from "./status.ts";
46
47
 
47
48
  const pkgPath = resolve(import.meta.dir, "../../package.json");
@@ -555,7 +556,7 @@ export function renderAgentPanel(
555
556
  output += `${CURSOR.cursorTo(startRow, 1)}${headerLine}${headerPadding}${dimBox.vertical}\n`;
556
557
 
557
558
  // Column headers
558
- const colStr = `${dimBox.vertical} St Name Capability State Task ID Duration Tmux `;
559
+ const colStr = `${dimBox.vertical} St Name Capability State Task ID Duration Live `;
559
560
  const colPadding = " ".repeat(
560
561
  Math.max(0, leftWidth - visibleLength(colStr) - visibleLength(dimBox.vertical)),
561
562
  );
@@ -595,10 +596,13 @@ export function renderAgentPanel(
595
596
  : now;
596
597
  const duration = formatDuration(endTime - new Date(agent.startedAt).getTime());
597
598
  const durationPadded = pad(duration, 9);
598
- const tmuxAlive = data.status.tmuxSessions.some((s) => s.name === agent.tmuxSession);
599
- const tmuxDot = tmuxAlive ? color.green(">") : color.red("x");
599
+ const isHeadless = agent.tmuxSession === "" && agent.pid !== null;
600
+ const alive = isHeadless
601
+ ? agent.pid !== null && isProcessAlive(agent.pid)
602
+ : data.status.tmuxSessions.some((s) => s.name === agent.tmuxSession);
603
+ const aliveDot = alive ? color.green(">") : color.red("x");
600
604
 
601
- const lineContent = `${dimBox.vertical} ${stateColorFn(icon)} ${name} ${capability} ${stateColorFn(state)} ${taskId} ${durationPadded} ${tmuxDot} `;
605
+ const lineContent = `${dimBox.vertical} ${stateColorFn(icon)} ${name} ${capability} ${stateColorFn(state)} ${taskId} ${durationPadded} ${aliveDot} `;
602
606
  const linePadding = " ".repeat(
603
607
  Math.max(0, leftWidth - visibleLength(lineContent) - visibleLength(dimBox.vertical)),
604
608
  );
@@ -469,6 +469,10 @@ describe("feedCommand", () => {
469
469
  "spawn",
470
470
  "error",
471
471
  "custom",
472
+ "turn_start",
473
+ "turn_end",
474
+ "progress",
475
+ "result",
472
476
  ] as const;
473
477
  for (const eventType of eventTypes) {
474
478
  store.insert(
@@ -494,6 +498,10 @@ describe("feedCommand", () => {
494
498
  expect(out).toContain("SPAWN");
495
499
  expect(out).toContain("ERROR");
496
500
  expect(out).toContain("CUSTM");
501
+ expect(out).toContain("TURN+");
502
+ expect(out).toContain("TURN-");
503
+ expect(out).toContain("PROG ");
504
+ expect(out).toContain("RSULT");
497
505
  });
498
506
  });
499
507
 
@@ -18,7 +18,7 @@ import { createMetricsStore } from "../metrics/store.ts";
18
18
  import { createSessionStore } from "../sessions/store.ts";
19
19
  import { cleanupTempDir } from "../test-helpers.ts";
20
20
  import type { InsertEvent, SessionMetrics } from "../types.ts";
21
- import { gatherInspectData, inspectCommand } from "./inspect.ts";
21
+ import { gatherInspectData, inspectCommand, printInspectData } from "./inspect.ts";
22
22
 
23
23
  /** Helper to create an InsertEvent with sensible defaults. */
24
24
  function makeEvent(overrides: Partial<InsertEvent> = {}): InsertEvent {
@@ -565,6 +565,161 @@ describe("inspectCommand", () => {
565
565
  });
566
566
  });
567
567
 
568
+ // === Headless agent support ===
569
+
570
+ describe("headless agent support", () => {
571
+ test("gatherInspectData skips tmux capture for headless agents (empty tmuxSession)", async () => {
572
+ const overstoryDir = join(tempDir, ".overstory");
573
+ const sessionsDbPath = join(overstoryDir, "sessions.db");
574
+ const store = createSessionStore(sessionsDbPath);
575
+ store.upsert({
576
+ id: "sess-h1",
577
+ agentName: "headless-agent",
578
+ capability: "builder",
579
+ worktreePath: "/tmp/wt",
580
+ branchName: "overstory/headless/task-1",
581
+ taskId: "overstory-h01",
582
+ tmuxSession: "", // headless
583
+ state: "working",
584
+ pid: process.pid,
585
+ parentAgent: null,
586
+ depth: 0,
587
+ runId: null,
588
+ startedAt: new Date().toISOString(),
589
+ lastActivity: new Date().toISOString(),
590
+ escalationLevel: 0,
591
+ stalledSince: null,
592
+ transcriptPath: null,
593
+ });
594
+ store.close();
595
+
596
+ // noTmux=false but tmuxSession="" — should skip tmux capture without error
597
+ const data = await gatherInspectData(tempDir, "headless-agent", { noTmux: false });
598
+ // tmuxOutput is null (no tmux) and no events yet → no fallback either
599
+ expect(data.session.agentName).toBe("headless-agent");
600
+ expect(data.session.tmuxSession).toBe("");
601
+ // tmuxOutput may be null (no events) or a string (fallback) — must not throw
602
+ });
603
+
604
+ test("gatherInspectData provides event-based output for headless agents with tool calls", async () => {
605
+ const overstoryDir = join(tempDir, ".overstory");
606
+ const sessionsDbPath = join(overstoryDir, "sessions.db");
607
+ const eventsDbPath = join(overstoryDir, "events.db");
608
+
609
+ const store = createSessionStore(sessionsDbPath);
610
+ store.upsert({
611
+ id: "sess-h2",
612
+ agentName: "headless-events",
613
+ capability: "builder",
614
+ worktreePath: "/tmp/wt",
615
+ branchName: "overstory/headless/task-2",
616
+ taskId: "overstory-h02",
617
+ tmuxSession: "", // headless
618
+ state: "working",
619
+ pid: process.pid,
620
+ parentAgent: null,
621
+ depth: 0,
622
+ runId: null,
623
+ startedAt: new Date().toISOString(),
624
+ lastActivity: new Date().toISOString(),
625
+ escalationLevel: 0,
626
+ stalledSince: null,
627
+ transcriptPath: null,
628
+ });
629
+ store.close();
630
+
631
+ const eventStore = createEventStore(eventsDbPath);
632
+ eventStore.insert(
633
+ makeEvent({ agentName: "headless-events", toolName: "Read", toolDurationMs: 50 }),
634
+ );
635
+ eventStore.insert(
636
+ makeEvent({ agentName: "headless-events", toolName: "Edit", toolDurationMs: 100 }),
637
+ );
638
+ eventStore.close();
639
+
640
+ const data = await gatherInspectData(tempDir, "headless-events", { noTmux: false });
641
+
642
+ // Should have fallback output
643
+ expect(data.tmuxOutput).not.toBeNull();
644
+ expect(data.tmuxOutput).toContain("Headless agent");
645
+ expect(data.tmuxOutput).toContain("Read");
646
+ });
647
+
648
+ test("printInspectData shows PID instead of tmux session for headless agents", () => {
649
+ const data = {
650
+ session: {
651
+ id: "sess-h3",
652
+ agentName: "headless-display",
653
+ capability: "builder",
654
+ worktreePath: "/tmp/wt",
655
+ branchName: "overstory/headless/task-3",
656
+ taskId: "overstory-h03",
657
+ tmuxSession: "", // headless
658
+ state: "working" as const,
659
+ pid: 99999,
660
+ parentAgent: null,
661
+ depth: 0,
662
+ runId: null,
663
+ startedAt: new Date().toISOString(),
664
+ lastActivity: new Date().toISOString(),
665
+ escalationLevel: 0,
666
+ stalledSince: null,
667
+ transcriptPath: null,
668
+ },
669
+ timeSinceLastActivity: 5000,
670
+ recentToolCalls: [],
671
+ currentFile: null,
672
+ toolStats: [],
673
+ tokenUsage: null,
674
+ tmuxOutput: null,
675
+ };
676
+
677
+ printInspectData(data);
678
+
679
+ const out = output();
680
+ expect(out).toContain("Process: PID");
681
+ expect(out).toContain("99999");
682
+ expect(out).toContain("headless");
683
+ expect(out).not.toContain("Tmux:");
684
+ });
685
+
686
+ test("printInspectData shows Recent Activity header for headless agents with tmuxOutput", () => {
687
+ const data = {
688
+ session: {
689
+ id: "sess-h4",
690
+ agentName: "headless-activity",
691
+ capability: "builder",
692
+ worktreePath: "/tmp/wt",
693
+ branchName: "overstory/headless/task-4",
694
+ taskId: "overstory-h04",
695
+ tmuxSession: "", // headless
696
+ state: "working" as const,
697
+ pid: 99998,
698
+ parentAgent: null,
699
+ depth: 0,
700
+ runId: null,
701
+ startedAt: new Date().toISOString(),
702
+ lastActivity: new Date().toISOString(),
703
+ escalationLevel: 0,
704
+ stalledSince: null,
705
+ transcriptPath: null,
706
+ },
707
+ timeSinceLastActivity: 5000,
708
+ recentToolCalls: [],
709
+ currentFile: null,
710
+ toolStats: [],
711
+ tokenUsage: null,
712
+ tmuxOutput: "[Headless agent — showing recent tool events]",
713
+ };
714
+
715
+ printInspectData(data);
716
+
717
+ const out = output();
718
+ expect(out).toContain("Recent Activity (headless)");
719
+ expect(out).not.toContain("Live Tmux Output");
720
+ });
721
+ });
722
+
568
723
  // === Human-readable output ===
569
724
 
570
725
  describe("human-readable output", () => {
@@ -193,13 +193,24 @@ export async function gatherInspectData(
193
193
  }
194
194
  }
195
195
 
196
- // tmux capture
196
+ // tmux capture (skipped for headless agents where tmuxSession is empty)
197
197
  let tmuxOutput: string | null = null;
198
198
  if (!opts.noTmux && session.tmuxSession) {
199
199
  const lines = opts.tmuxLines ?? 30;
200
200
  tmuxOutput = await captureTmux(session.tmuxSession, lines);
201
201
  }
202
202
 
203
+ // Headless fallback: show recent events as live output when no tmux
204
+ if (!tmuxOutput && session.tmuxSession === "" && recentToolCalls.length > 0) {
205
+ const lines: string[] = ["[Headless agent — showing recent tool events]", ""];
206
+ for (const call of recentToolCalls.slice(0, 15)) {
207
+ const time = new Date(call.timestamp).toLocaleTimeString();
208
+ const dur = call.durationMs !== null ? `${call.durationMs}ms` : "pending";
209
+ lines.push(` [${time}] ${call.toolName.padEnd(15)} ${dur}`);
210
+ }
211
+ tmuxOutput = lines.join("\n");
212
+ }
213
+
203
214
  return {
204
215
  session,
205
216
  timeSinceLastActivity,
@@ -233,7 +244,11 @@ export function printInspectData(data: InspectData): void {
233
244
  w(`Parent: ${accent(session.parentAgent)} (depth: ${session.depth})\n`);
234
245
  }
235
246
  w(`Started: ${session.startedAt}\n`);
236
- w(`Tmux: ${accent(session.tmuxSession)}\n`);
247
+ if (session.tmuxSession) {
248
+ w(`Tmux: ${accent(session.tmuxSession)}\n`);
249
+ } else if (session.pid !== null) {
250
+ w(`Process: PID ${accent(String(session.pid))} (headless)\n`);
251
+ }
237
252
  w("\n");
238
253
 
239
254
  // Current file
@@ -287,9 +302,9 @@ export function printInspectData(data: InspectData): void {
287
302
  w("\n");
288
303
  }
289
304
 
290
- // tmux output
305
+ // tmux output (or headless fallback)
291
306
  if (data.tmuxOutput) {
292
- w("Live Tmux Output\n");
307
+ w(data.session.tmuxSession ? "Live Tmux Output\n" : "Recent Activity (headless)\n");
293
308
  w(`${separator()}\n`);
294
309
  w(`${data.tmuxOutput}\n`);
295
310
  w(`${separator()}\n`);
@@ -701,6 +701,10 @@ describe("replayCommand", () => {
701
701
  "spawn",
702
702
  "error",
703
703
  "custom",
704
+ "turn_start",
705
+ "turn_end",
706
+ "progress",
707
+ "result",
704
708
  ] as const;
705
709
  for (const eventType of eventTypes) {
706
710
  store.insert(
@@ -724,6 +728,10 @@ describe("replayCommand", () => {
724
728
  expect(out).toContain("SPAWN");
725
729
  expect(out).toContain("ERROR");
726
730
  expect(out).toContain("CUSTOM");
731
+ expect(out).toContain("TURN START");
732
+ expect(out).toContain("TURN END");
733
+ expect(out).toContain("PROGRESS");
734
+ expect(out).toContain("RESULT");
727
735
  });
728
736
 
729
737
  test("long data values are truncated", async () => {