pi-crew 0.7.5 → 0.7.6

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 (51) hide show
  1. package/CHANGELOG.md +51 -0
  2. package/README.md +11 -11
  3. package/docs/commands-reference.md +14 -10
  4. package/docs/troubleshooting.md +131 -0
  5. package/docs/usage.md +9 -4
  6. package/package.json +1 -1
  7. package/src/config/config.ts +11 -4
  8. package/src/extension/action-suggestions.ts +71 -0
  9. package/src/extension/context-status-injection.ts +32 -1
  10. package/src/extension/register.ts +71 -65
  11. package/src/extension/team-tool/api.ts +3 -2
  12. package/src/extension/team-tool/cancel.ts +5 -4
  13. package/src/extension/team-tool/explain.ts +2 -1
  14. package/src/extension/team-tool/failure-patterns.ts +124 -0
  15. package/src/extension/team-tool/inspect.ts +10 -6
  16. package/src/extension/team-tool/lifecycle-actions.ts +5 -4
  17. package/src/extension/team-tool/respond.ts +4 -3
  18. package/src/extension/team-tool/run-not-found.ts +54 -0
  19. package/src/extension/team-tool/run.ts +26 -4
  20. package/src/extension/team-tool/status.ts +58 -4
  21. package/src/extension/team-tool.ts +5 -3
  22. package/src/runtime/async-runner.ts +7 -0
  23. package/src/runtime/background-runner.ts +7 -1
  24. package/src/runtime/chain-parser.ts +13 -5
  25. package/src/runtime/checkpoint.ts +13 -1
  26. package/src/runtime/child-pi.ts +9 -1
  27. package/src/runtime/live-session-runtime.ts +15 -1
  28. package/src/runtime/parent-guard.ts +2 -2
  29. package/src/runtime/stale-reconciler.ts +8 -3
  30. package/src/runtime/task-runner.ts +10 -1
  31. package/src/runtime/team-runner.ts +19 -2
  32. package/src/runtime/verification-gates.ts +21 -1
  33. package/src/schema/team-tool-schema.ts +9 -0
  34. package/src/state/blob-store.ts +12 -10
  35. package/src/state/event-log-rotation.ts +114 -93
  36. package/src/state/event-log.ts +79 -20
  37. package/src/state/health-store.ts +6 -1
  38. package/src/state/locks.ts +66 -16
  39. package/src/state/state-store.ts +14 -1
  40. package/src/ui/card-colors.ts +7 -3
  41. package/src/ui/dashboard-panes/agents-pane.ts +15 -2
  42. package/src/ui/live-duration.ts +58 -0
  43. package/src/ui/tool-render.ts +7 -11
  44. package/src/ui/tool-renderers/index.ts +6 -3
  45. package/src/ui/widget/widget-formatters.ts +2 -13
  46. package/src/utils/fs-watch.ts +11 -60
  47. package/src/utils/run-watcher-registry.ts +164 -0
  48. package/src/workflows/discover-workflows.ts +2 -1
  49. package/src/workflows/workflow-config.ts +5 -0
  50. package/src/runtime/dynamic-script-runner.ts +0 -497
  51. package/src/runtime/sandbox.ts +0 -335
@@ -24,6 +24,7 @@ import { resolveRealContainedPath } from "../../utils/safe-paths.ts";
24
24
  import type { PiTeamsToolResult } from "../tool-result.ts";
25
25
  import { locateRunCwd } from "../team-tool.ts";
26
26
  import { configRecord, result, type TeamContext } from "./context.ts";
27
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
27
28
 
28
29
  export function globMatch(value: string, pattern: string): boolean {
29
30
  // Prevent ReDoS: reject excessively long patterns
@@ -91,9 +92,9 @@ export async function handleApi(params: TeamToolParamsValue, ctx: TeamContext):
91
92
  }
92
93
  if (!params.runId) return result("API requires runId.", { action: "api", status: "error" }, true);
93
94
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
94
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "api", status: "error" }, true);
95
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "api", status: "error" }, true);
95
96
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
96
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "api", status: "error" }, true);
97
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "api", status: "error" }, true);
97
98
  if (operation === "read-manifest") {
98
99
  return result(JSON.stringify(loaded.manifest, null, 2), { action: "api", status: "ok", runId: loaded.manifest.runId, artifactsRoot: loaded.manifest.artifactsRoot });
99
100
  }
@@ -12,6 +12,7 @@ import { executeHook, appendHookEvent } from "../../hooks/registry.ts";
12
12
  import type { PiTeamsToolResult } from "../tool-result.ts";
13
13
  import { locateRunCwd } from "../team-tool.ts";
14
14
  import { result, type TeamContext } from "./context.ts";
15
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
15
16
  import { enforceDestructiveIntent, intentFromConfig } from "./intent-policy.ts";
16
17
  import { invalidateSnapshot, type CacheControlDeps } from "./cache-control.ts";
17
18
 
@@ -80,9 +81,9 @@ function cancelReasonFromParams(params: TeamToolParamsValue): CancellationReason
80
81
  export async function handleRetry(params: TeamToolParamsValue, ctx: TeamContext, deps?: CacheControlDeps): Promise<PiTeamsToolResult> {
81
82
  if (!params.runId) return result("Retry requires runId.", { action: "retry", status: "error" }, true);
82
83
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
83
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "retry", status: "error" }, true);
84
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "retry", status: "error" }, true);
84
85
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
85
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "retry", status: "error" }, true);
86
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "retry", status: "error" }, true);
86
87
 
87
88
  // Pre-lock ownership check: reject foreign-owned runs unless force is set
88
89
  const foreignRun = typeof loaded.manifest.ownerSessionId === "string" && loaded.manifest.ownerSessionId !== ctx.sessionId;
@@ -145,9 +146,9 @@ export async function handleCancel(params: TeamToolParamsValue, ctx: TeamContext
145
146
  if (intentError) return intentError;
146
147
  if (!params.runId) return result("Cancel requires runId.", { action: "cancel", status: "error" }, true);
147
148
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
148
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "cancel", status: "error" }, true);
149
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "cancel", status: "error" }, true);
149
150
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
150
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "cancel", status: "error" }, true);
151
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "cancel", status: "error" }, true);
151
152
 
152
153
  // Pre-lock ownership check: reject foreign-owned runs unless force is set
153
154
  const preCheck = abortOwned(loaded.manifest.runId, undefined, ctx, params.force);
@@ -1,4 +1,5 @@
1
1
  import * as fs from "node:fs";
2
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
2
3
  import * as path from "node:path";
3
4
  import { loadRunManifestById } from "../../state/state-store.ts";
4
5
  import type { TeamRunManifest, TeamTaskState } from "../../state/types.ts";
@@ -211,7 +212,7 @@ export function handleExplain(params: {
211
212
 
212
213
  const loaded = loadRunManifestById(cwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
213
214
  if (!loaded) {
214
- return result(`Run '${params.runId}' not found.`, { action: "explain", status: "error" }, true);
215
+ return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "explain", status: "error" }, true);
215
216
  }
216
217
 
217
218
  const { manifest, tasks } = loaded;
@@ -0,0 +1,124 @@
1
+ /**
2
+ * failure-patterns.ts — Group failed tasks by error similarity (Round 17 BS-4).
3
+ *
4
+ * Before this, a run with 8 failed tasks surfaced 8 separate raw error
5
+ * strings. The user had to mentally group them ("5 of these say 'model
6
+ * routing fallback failed'"). This module detects common failure patterns
7
+ * so `summary` can say "5 of 8 failures share root cause: X".
8
+ *
9
+ * Grouping strategy (cheap, deterministic, no ML):
10
+ * 1. Normalize: lowercase, collapse whitespace, strip task ids / run ids /
11
+ * absolute paths / numbers → a canonical "signature".
12
+ * 2. Bucket by signature. Buckets with >1 member are "common patterns".
13
+ * 3. Sort by frequency desc.
14
+ *
15
+ * Conservative: only buckets with >=2 members count as a pattern (a single
16
+ * failure is just itself). Returns [] when there are no repeated signatures.
17
+ */
18
+
19
+ export interface FailurePattern {
20
+ /** Canonical error signature used for grouping. */
21
+ signature: string;
22
+ /** A representative original error (the shortest variant) for display. */
23
+ representative: string;
24
+ /** Task ids that hit this pattern. */
25
+ taskIds: string[];
26
+ /** Count of failures in this bucket (== taskIds.length). */
27
+ count: number;
28
+ }
29
+
30
+ export interface FailurePatternInput {
31
+ id: string;
32
+ status: string;
33
+ error?: string;
34
+ }
35
+
36
+ /**
37
+ * Normalize an error string into a grouping signature.
38
+ * Exported for unit testing.
39
+ */
40
+ export function normalizeErrorSignature(error: string | undefined): string {
41
+ if (!error) return "(no error detail)";
42
+ let s = error.toLowerCase();
43
+ // Strip run ids (team_YYYYMMDDHHMMSS_xxxxxxxxxxxxxxxx)
44
+ s = s.replace(/team_\d{8,}_[a-z0-9]{12,}/g, "<run>");
45
+ // Strip task ids (01_explore, adaptive-03-executor, etc.)
46
+ s = s.replace(/\b(adaptive-)?\d{2,}[a-z0-9_-]+/g, "<task>");
47
+ // Strip absolute paths
48
+ s = s.replace(/\/(?:home|users|tmp|var|opt|root)[^\s'"]*/g, "<path>");
49
+ // Strip numbers (line numbers, counts, pids, ms durations)
50
+ s = s.replace(/\b\d+\b/g, "N");
51
+ // Collapse whitespace
52
+ s = s.replace(/\s+/g, " ").trim();
53
+ return s || "(no error detail)";
54
+ }
55
+
56
+ /**
57
+ * Group failed tasks by error-pattern similarity. Only groups with >=2
58
+ * members are returned (singletons are not "patterns"). Sorted by count desc.
59
+ *
60
+ * @param tasks the run's tasks (any with status 'failed'/'cancelled' are
61
+ * considered failures for aggregation purposes).
62
+ */
63
+ export function aggregateFailurePatterns(tasks: FailurePatternInput[]): FailurePattern[] {
64
+ const failed = tasks.filter(
65
+ (t) => t.status === "failed" || t.status === "cancelled",
66
+ );
67
+ if (failed.length === 0) return [];
68
+ const buckets = new Map<string, FailurePattern>();
69
+ for (const t of failed) {
70
+ const signature = normalizeErrorSignature(t.error);
71
+ const existing = buckets.get(signature);
72
+ if (existing) {
73
+ existing.taskIds.push(t.id);
74
+ existing.count += 1;
75
+ // Keep the shortest non-empty variant as representative (most readable).
76
+ if (t.error && (!existing.representative || t.error.length < existing.representative.length)) {
77
+ existing.representative = t.error;
78
+ }
79
+ } else {
80
+ buckets.set(signature, {
81
+ signature,
82
+ representative: t.error ?? "(no error detail)",
83
+ taskIds: [t.id],
84
+ count: 1,
85
+ });
86
+ }
87
+ }
88
+ // Only patterns with >=2 members (repeated root causes).
89
+ return [...buckets.values()]
90
+ .filter((b) => b.count >= 2)
91
+ .sort((a, b) => b.count - a.count);
92
+ }
93
+
94
+ /**
95
+ * Render failure patterns as human-readable lines for the `summary` action.
96
+ * Returns [] when there are no repeated patterns (so the caller can omit the
97
+ * section entirely).
98
+ *
99
+ * Example output:
100
+ * Common failure patterns (3 of 5 failures share 2 root causes):
101
+ * - [×3] model routing fallback failed: all 2 candidates exhausted
102
+ * tasks: 02_exec, 03_exec, 04_exec
103
+ * - [×2] EPERM: operation not permitted, rename
104
+ * tasks: 05_exec, 06_exec
105
+ */
106
+ export function formatFailurePatterns(tasks: FailurePatternInput[]): string[] {
107
+ const patterns = aggregateFailurePatterns(tasks);
108
+ if (patterns.length === 0) return [];
109
+ const failedCount = tasks.filter(
110
+ (t) => t.status === "failed" || t.status === "cancelled",
111
+ ).length;
112
+ const groupedCount = patterns.reduce((sum, p) => sum + p.count, 0);
113
+ const lines = [
114
+ `Common failure patterns (${groupedCount} of ${failedCount} failures share ${patterns.length} root cause${patterns.length === 1 ? "" : "s"}):`,
115
+ ];
116
+ for (const p of patterns) {
117
+ const rep = p.representative.length > 100 ? `${p.representative.slice(0, 99)}…` : p.representative;
118
+ lines.push(`- [×${p.count}] ${rep}`);
119
+ const shown = p.taskIds.slice(0, 6);
120
+ const more = p.taskIds.length > 6 ? `, +${p.taskIds.length - 6} more` : "";
121
+ lines.push(` tasks: ${shown.join(", ")}${more}`);
122
+ }
123
+ return lines;
124
+ }
@@ -5,13 +5,15 @@ import { aggregateUsage, formatUsage, formatCostReport } from "../../state/usage
5
5
  import type { PiTeamsToolResult } from "../tool-result.ts";
6
6
  import { locateRunCwd } from "../team-tool.ts";
7
7
  import { result, type TeamContext } from "./context.ts";
8
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
9
+ import { formatFailurePatterns } from "./failure-patterns.ts";
8
10
 
9
11
  export function handleEvents(params: TeamToolParamsValue, ctx: TeamContext): PiTeamsToolResult {
10
12
  if (!params.runId) return result("Events requires runId.", { action: "events", status: "error" }, true);
11
13
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
12
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "events", status: "error" }, true);
14
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "events", status: "error" }, true);
13
15
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
14
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "events", status: "error" }, true);
16
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "events", status: "error" }, true);
15
17
  const events = readEvents(loaded.manifest.eventsPath);
16
18
  const lines = [`Events for ${loaded.manifest.runId}:`, ...(events.length ? events.map((event) => `${event.time} ${event.type}${event.taskId ? ` ${event.taskId}` : ""}${event.message ? `: ${event.message}` : ""}${event.data ? ` ${JSON.stringify(event.data)}` : ""}`) : ["(none)"])];
17
19
  return result(lines.join("\n"), { action: "events", status: "ok", runId: loaded.manifest.runId, artifactsRoot: loaded.manifest.artifactsRoot });
@@ -20,9 +22,9 @@ export function handleEvents(params: TeamToolParamsValue, ctx: TeamContext): PiT
20
22
  export function handleArtifacts(params: TeamToolParamsValue, ctx: TeamContext): PiTeamsToolResult {
21
23
  if (!params.runId) return result("Artifacts requires runId.", { action: "artifacts", status: "error" }, true);
22
24
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
23
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "artifacts", status: "error" }, true);
25
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "artifacts", status: "error" }, true);
24
26
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
25
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "artifacts", status: "error" }, true);
27
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "artifacts", status: "error" }, true);
26
28
  const lines = [`Artifacts for ${loaded.manifest.runId}:`, ...(loaded.manifest.artifacts.length ? loaded.manifest.artifacts.map((artifact) => `- ${artifact.kind}: ${artifact.path}${artifact.sizeBytes !== undefined ? ` (${artifact.sizeBytes} bytes)` : ""}${artifact.contentHash ? ` sha256=${artifact.contentHash.slice(0, 12)}` : ""}`) : ["- (none)"])];
27
29
  return result(lines.join("\n"), { action: "artifacts", status: "ok", runId: loaded.manifest.runId, artifactsRoot: loaded.manifest.artifactsRoot });
28
30
  }
@@ -30,10 +32,11 @@ export function handleArtifacts(params: TeamToolParamsValue, ctx: TeamContext):
30
32
  export function handleSummary(params: TeamToolParamsValue, ctx: TeamContext): PiTeamsToolResult {
31
33
  if (!params.runId) return result("Summary requires runId.", { action: "summary", status: "error" }, true);
32
34
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
33
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "summary", status: "error" }, true);
35
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "summary", status: "error" }, true);
34
36
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
35
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "summary", status: "error" }, true);
37
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "summary", status: "error" }, true);
36
38
  const usage = aggregateUsage(loaded.tasks);
39
+ const failurePatternLines = formatFailurePatterns(loaded.tasks);
37
40
  const lines = [
38
41
  `Summary for ${loaded.manifest.runId}`,
39
42
  `Status: ${loaded.manifest.status}`,
@@ -43,6 +46,7 @@ export function handleSummary(params: TeamToolParamsValue, ctx: TeamContext): Pi
43
46
  `Usage: ${formatUsage(usage)}`,
44
47
  "",
45
48
  formatCostReport(loaded.tasks),
49
+ ...(failurePatternLines.length > 0 ? ["", ...failurePatternLines] : []),
46
50
  "",
47
51
  "Tasks:",
48
52
  ...loaded.tasks.map((task) => `- ${task.id}: ${task.status} (${task.role} -> ${task.agent})${task.error ? ` - ${task.error}` : ""}`),
@@ -9,6 +9,7 @@ import { importRunBundle } from "../run-import.ts";
9
9
  import { pruneFinishedRuns } from "../run-maintenance.ts";
10
10
  import type { PiTeamsToolResult } from "../tool-result.ts";
11
11
  import { configRecord, result, type TeamContext } from "./context.ts";
12
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
12
13
  import { enforceDestructiveIntent, intentFromConfig } from "./intent-policy.ts";
13
14
  import { executeHook, appendHookEvent } from "../../hooks/registry.ts";
14
15
  import { resolveRealContainedPath } from "../../utils/safe-paths.ts";
@@ -18,7 +19,7 @@ import * as path from "node:path";
18
19
  export function handleWorktrees(params: TeamToolParamsValue, ctx: TeamContext): PiTeamsToolResult {
19
20
  if (!params.runId) return result("Worktrees requires runId.", { action: "worktrees", status: "error" }, true);
20
21
  const loaded = loadRunManifestById(ctx.cwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
21
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "worktrees", status: "error" }, true);
22
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "worktrees", status: "error" }, true);
22
23
  const withWorktrees = loaded.tasks.filter((task) => task.worktree);
23
24
  const lines = [`Worktrees for ${loaded.manifest.runId}:`, ...(withWorktrees.length ? withWorktrees.map((task) => `- ${task.id}: ${task.worktree!.path} branch=${task.worktree!.branch} reused=${task.worktree!.reused ? "true" : "false"}`) : ["- (none)"])];
24
25
  return result(lines.join("\n"), { action: "worktrees", status: "ok", runId: loaded.manifest.runId, artifactsRoot: loaded.manifest.artifactsRoot });
@@ -47,7 +48,7 @@ export function handleImport(params: TeamToolParamsValue, ctx: TeamContext): PiT
47
48
  export async function handleExport(params: TeamToolParamsValue, ctx: TeamContext): Promise<PiTeamsToolResult> {
48
49
  if (!params.runId) return result("Export requires runId.", { action: "export", status: "error" }, true);
49
50
  const loaded = loadRunManifestById(ctx.cwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
50
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "export", status: "error" }, true);
51
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "export", status: "error" }, true);
51
52
 
52
53
  // SECURITY: Ownership check — only the owner session may export a run.
53
54
  // Foreign-run export requires confirm: true (explicit user intent).
@@ -96,7 +97,7 @@ export async function handleForget(params: TeamToolParamsValue, ctx: TeamContext
96
97
  if (!params.runId) return result("Forget requires runId.", { action: "forget", status: "error" }, true);
97
98
  if (!params.confirm) return result("forget requires confirm: true.", { action: "forget", status: "error" }, true);
98
99
  const loaded = loadRunManifestById(ctx.cwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
99
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "forget", status: "error" }, true);
100
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "forget", status: "error" }, true);
100
101
 
101
102
  // Ownership check — prevent cross-session deletion unless force is set
102
103
  const foreignRun = typeof loaded.manifest.ownerSessionId === "string" && loaded.manifest.ownerSessionId !== ctx.sessionId;
@@ -126,7 +127,7 @@ export async function handleCleanup(params: TeamToolParamsValue, ctx: TeamContex
126
127
  if (intentError) return intentError;
127
128
  if (!params.runId) return result("Cleanup requires runId.", { action: "cleanup", status: "error" }, true);
128
129
  const loaded = loadRunManifestById(ctx.cwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
129
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "cleanup", status: "error" }, true);
130
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "cleanup", status: "error" }, true);
130
131
 
131
132
  // Ownership check — prevent cross-session worktree cleanup unless force is set
132
133
  const foreignRun = typeof loaded.manifest.ownerSessionId === "string" && loaded.manifest.ownerSessionId !== ctx.sessionId;
@@ -8,6 +8,7 @@ import { logInternalError } from "../../utils/internal-error.ts";
8
8
  import type { PiTeamsToolResult } from "../tool-result.ts";
9
9
  import { locateRunCwd } from "../team-tool.ts";
10
10
  import { result, type TeamContext } from "./context.ts";
11
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
11
12
 
12
13
  /**
13
14
  * Handle `respond` action: send a message to a waiting (interactive) task.
@@ -19,13 +20,13 @@ export function handleRespond(params: TeamToolParamsValue, ctx: TeamContext): Pi
19
20
  if (!params.message && !params.taskId) return result("Respond requires taskId and/or message.", { action: "respond", status: "error" }, true);
20
21
 
21
22
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
22
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "respond", status: "error" }, true);
23
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "respond", status: "error" }, true);
23
24
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
24
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "respond", status: "error" }, true);
25
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "respond", status: "error" }, true);
25
26
 
26
27
  return withRunLockSync(loaded.manifest, () => {
27
28
  const fresh = loadRunManifestById(loaded.manifest.cwd, params.runId!); // NOTE: inside withRunLockSync - consistent read
28
- if (!fresh) return result(`Run '${params.runId}' not found.`, { action: "respond", status: "error" }, true);
29
+ if (!fresh) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "respond", status: "error" }, true);
29
30
  const foreignRun = typeof fresh.manifest.ownerSessionId === "string" && fresh.manifest.ownerSessionId !== ctx.sessionId;
30
31
  if (foreignRun && !params.force) return result(`Run ${fresh.manifest.runId} belongs to another session. Use force: true to override.`, { action: "respond", status: "error", runId: fresh.manifest.runId }, true);
31
32
 
@@ -0,0 +1,54 @@
1
+ /**
2
+ * run-not-found.ts — Centralized "Run not found" error helper (DX: F2).
3
+ *
4
+ * Round 16 DX audit found that a stale/typo'd runId hits a blank
5
+ * "Run '<id>' not found." wall in 8+ handlers (status, resume, steer, export,
6
+ * forget, cleanup, invalidate, worktrees, events, artifacts). The run IDs are
7
+ * long (`team_20260615173318_b9c8fe49a74e0760`), so typos/truncation are
8
+ * near-certain for new users — yet `team list` (which shows recent runs) is
9
+ * never suggested.
10
+ *
11
+ * This module centralizes the message + recovery hint so every handler stays
12
+ * consistent and the hint never drifts.
13
+ */
14
+
15
+ import { result, type TeamContext } from "./context.ts";
16
+ import type { TeamToolDetails } from "../team-tool-types.ts";
17
+
18
+ /** Recovery hint appended to every "Run not found" message. */
19
+ export const RUN_NOT_FOUND_HINT =
20
+ "\n\nTip: run action='list' to see recent runs and their IDs.";
21
+
22
+ /**
23
+ * Build the standard "Run not found" error result with a recovery hint.
24
+ *
25
+ * @param runId the (missing/typo'd) run id the caller passed
26
+ * @param action the action that was attempted (for the details.action field)
27
+ */
28
+ export function runNotFound(runId: string, action: string): ReturnType<typeof result> {
29
+ return result(
30
+ `Run '${runId}' not found.${RUN_NOT_FOUND_HINT}`,
31
+ { action, status: "error" } satisfies TeamToolDetails,
32
+ true,
33
+ );
34
+ }
35
+
36
+ /**
37
+ * Helper: resolve a runId to its cwd, returning a runNotFound() result when
38
+ * missing. Reduces the boilerplate `locateRunCwd → if (!runCwd) return ...`
39
+ * duplicated across handlers.
40
+ */
41
+ export function resolveRunOrNotFound(
42
+ runId: string,
43
+ action: string,
44
+ cwd: string,
45
+ locate: (runId: string, cwd: string) => string | undefined,
46
+ ): { kind: "found"; runCwd: string } | { kind: "notfound"; result: ReturnType<typeof result> } {
47
+ const runCwd = locate(runId, cwd);
48
+ if (!runCwd) return { kind: "notfound", result: runNotFound(runId, action) };
49
+ return { kind: "found", runCwd };
50
+ }
51
+
52
+ // Re-export TeamContext so callers importing this helper don't need a second
53
+ // import line — keeps the diff in each handler to a single import swap.
54
+ export type { TeamContext };
@@ -184,13 +184,17 @@ export async function handleRun(params: TeamToolParamsValue, ctx: TeamContext):
184
184
  // connecting PipelineRunner to the actual team execution system
185
185
  const stageInfo = pipelineWorkflow.stages.map((s) => `- ${s.name} (${s.team})`).join("\n");
186
186
  return result([
187
- `Pipeline workflow: ${workflow.name}`,
187
+ `Pipeline workflow '${workflow.name}' is not yet wired into the team execution system.`,
188
188
  `Goal: ${goal}`,
189
- `Stages (${pipelineWorkflow.stages.length}):`,
189
+ `Defined stages (${pipelineWorkflow.stages.length}):`,
190
190
  stageInfo,
191
191
  "",
192
- "Pipeline execution is available via the PipelineRunner API.",
193
- "Full CLI integration requires connecting to the team execution system.",
192
+ "To actually run work right now, use a supported workflow instead:",
193
+ " - action='run' workflow='default' (explore plan execute verify)",
194
+ " - action='run' workflow='implementation' (adaptive, parallel specialists)",
195
+ " - action='run' workflow='research' (explore → analyze → write)",
196
+ "",
197
+ "Run action='list' resource='workflow' to see all available workflows.",
194
198
  ].join("\n"), { action: "run", status: "ok" }, false);
195
199
  }
196
200
 
@@ -219,6 +223,24 @@ export async function handleRun(params: TeamToolParamsValue, ctx: TeamContext):
219
223
  registerActiveRun(updatedManifest);
220
224
 
221
225
  const loadedConfig = loadConfig(resolvedCtx.cwd);
226
+ // DX (Round 16 F4): surface config errors/warnings instead of silently
227
+ // proceeding with defaults. Non-blocking: emit a config.warning event so
228
+ // it shows in the run timeline and status, and log it. A malformed config
229
+ // (bad JSON / wrong types) should not be a silent no-op — doctor/config
230
+ // actions already surface these; run should too.
231
+ const configIssues = [
232
+ ...(loadedConfig.error ? [`Config error: ${loadedConfig.error}`] : []),
233
+ ...(loadedConfig.warnings ?? []),
234
+ ];
235
+ if (configIssues.length > 0) {
236
+ void appendEventAsync(updatedManifest.eventsPath, {
237
+ type: "config.warning",
238
+ runId: updatedManifest.runId,
239
+ message: `Loaded config from ${loadedConfig.path || "(defaults)"} with ${configIssues.length} issue(s): ${configIssues.join("; ")}`,
240
+ data: { error: loadedConfig.error, warnings: loadedConfig.warnings, path: loadedConfig.path },
241
+ }).catch((error) => logInternalError("team-tool.run.configWarning", error, `runId=${updatedManifest.runId}`));
242
+ logInternalError("team-tool.run.configWarning", new Error(`config issues: ${configIssues.join("; ")}`), `runId=${updatedManifest.runId} path=${loadedConfig.path ?? "(defaults)"}`);
243
+ }
222
244
  const executedConfig = effectiveRunConfig(loadedConfig.config, params.config);
223
245
  const runtime = await resolveCrewRuntime(executedConfig);
224
246
  const runtimeResolution = runtimeResolutionState(runtime);
@@ -3,24 +3,31 @@ import type { TeamToolParamsValue } from "../../schema/team-tool-schema.ts";
3
3
  import { appendEvent, readEvents } from "../../state/event-log.ts";
4
4
  import { readDeliveryState, readMailbox } from "../../state/mailbox.ts";
5
5
  import { loadRunManifestById, updateRunStatus, saveRunTasks } from "../../state/state-store.ts";
6
- import { aggregateUsage, formatUsage } from "../../state/usage.ts";
6
+ import { aggregateUsage, formatUsage, formatCost } from "../../state/usage.ts";
7
7
  import { applyAttentionState, formatActivityAge, resolveCrewControlConfig } from "../../runtime/agent-control.ts";
8
8
  import { readCrewAgents } from "../../runtime/crew-agent-records.ts";
9
9
  import { checkProcessLiveness, isActiveRunStatus } from "../../runtime/process-status.ts";
10
10
  import { formatTaskGraphLines, waitingReason } from "../../runtime/task-display.ts";
11
+ import { computePhaseProgress } from "../../runtime/phase-progress.ts";
12
+ import { formatDuration } from "../../ui/tool-render.ts";
11
13
  import { verifyTaskCompletion } from "../../runtime/completion-guard.ts";
12
14
  import { evaluateRunEffectiveness } from "../../runtime/effectiveness.ts";
13
15
  import type { PiTeamsToolResult } from "../tool-result.ts";
14
16
  import { locateRunCwd } from "../team-tool.ts";
15
17
  import { result, type TeamContext } from "./context.ts";
18
+ import { RUN_NOT_FOUND_HINT } from "./run-not-found.ts";
16
19
 
17
20
  export function handleStatus(params: TeamToolParamsValue, ctx: TeamContext): PiTeamsToolResult {
18
21
  if (!params.runId) return result("Status requires runId.", { action: "status", status: "error" }, true);
19
22
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
20
- if (!runCwd) return result(`Run '${params.runId}' not found.`, { action: "status", status: "error" }, true);
23
+ if (!runCwd) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "status", status: "error" }, true);
21
24
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
22
- if (!loaded) return result(`Run '${params.runId}' not found.`, { action: "status", status: "error" }, true);
25
+ if (!loaded) return result(`Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`, { action: "status", status: "error" }, true);
23
26
  let { manifest, tasks } = loaded;
27
+ // DX (Round 16 F3): compact status mode. Default = full (backward compatible).
28
+ // details=false gives a tight summary (status, goal, counts, failed/attention
29
+ // errors) for quick checks without 40 lines of dense key=value noise.
30
+ const fullDetails = params.details !== false;
24
31
  let asyncLivenessLine: string | undefined;
25
32
  if (manifest.async) {
26
33
  const asyncState = manifest.async;
@@ -35,6 +42,7 @@ export function handleStatus(params: TeamToolParamsValue, ctx: TeamContext): PiT
35
42
  }
36
43
  const counts = new Map<string, number>();
37
44
  for (const task of tasks) counts.set(task.status, (counts.get(task.status) ?? 0) + 1);
45
+ const phaseProgress = computePhaseProgress(tasks);
38
46
  const allEvents = readEvents(manifest.eventsPath);
39
47
  const events = allEvents.slice(-8);
40
48
  const attentionByTask = new Map(allEvents.filter((event) => event.type === "task.attention" && event.taskId).map((event) => [event.taskId!, event]));
@@ -62,12 +70,13 @@ export function handleStatus(params: TeamToolParamsValue, ctx: TeamContext): PiT
62
70
  const activeAgents = crewAgents.filter((agent) => agent.status === "running");
63
71
  const completedAgents = crewAgents.filter((agent) => agent.status !== "running");
64
72
  const waitingTasks = tasks.filter((task) => task.status === "queued" || task.status === "waiting");
65
- const agentLine = (agent: typeof crewAgents[number]): string => `- ${agent.id} [${agent.status}] ${agent.role} -> ${agent.agent} runtime=${agent.runtime}${agent.model ? ` model=${agent.model}` : ""}${agent.usage ? ` usage=${formatUsage(agent.usage)}` : ""}${agent.progress?.activityState ? ` activityState=${agent.progress.activityState}` : ""}${formatActivityAge(agent) ? ` activity=${formatActivityAge(agent)}` : ""}${agent.progress?.currentTool ? ` tool=${agent.progress.currentTool}` : ""}${agent.toolUses ? ` tools=${agent.toolUses}` : ""}${!agent.usage && agent.progress?.tokens ? ` tokens=${agent.progress.tokens}` : ""}${agent.progress?.turns ? ` turns=${agent.progress.turns}` : ""}${agent.jsonEvents !== undefined ? ` jsonEvents=${agent.jsonEvents}` : ""}${agent.outputPath ? ` output=${agent.outputPath}` : ""}${agent.transcriptPath ? ` transcript=${agent.transcriptPath}` : ""}${agent.statusPath ? ` status=${agent.statusPath}` : ""}${agent.error ? ` error=${agent.error}` : ""}`;
73
+ const agentLine = (agent: typeof crewAgents[number]): string => `- ${agent.id} [${agent.status}] ${agent.role} -> ${agent.agent} runtime=${agent.runtime}${agent.model ? ` model=${agent.model}` : ""}${agent.usage ? ` usage=${formatUsage(agent.usage)}` : ""}${agent.usage?.cost ? ` cost=${formatCost(agent.usage.cost)}` : ""}${agent.progress?.activityState ? ` activityState=${agent.progress.activityState}` : ""}${formatActivityAge(agent) ? ` activity=${formatActivityAge(agent)}` : ""}${agent.progress?.currentTool ? ` tool=${agent.progress.currentTool}` : ""}${agent.toolUses ? ` tools=${agent.toolUses}` : ""}${!agent.usage && agent.progress?.tokens ? ` tokens=${agent.progress.tokens}` : ""}${agent.progress?.turns ? ` turns=${agent.progress.turns}` : ""}${agent.jsonEvents !== undefined ? ` jsonEvents=${agent.jsonEvents}` : ""}${agent.outputPath ? ` output=${agent.outputPath}` : ""}${agent.transcriptPath ? ` transcript=${agent.transcriptPath}` : ""}${agent.statusPath ? ` status=${agent.statusPath}` : ""}${agent.error ? ` error=${agent.error}` : ""}`;
66
74
  const lines = [
67
75
  `Run: ${manifest.runId}`,
68
76
  `Team: ${manifest.team}`,
69
77
  `Workflow: ${manifest.workflow ?? "(none)"}`,
70
78
  `Status: ${manifest.status}`,
79
+ `Progress: ${phaseProgress.overallPercentage}% (~${formatDuration(phaseProgress.estimatedRemainingMs)} remaining)`,
71
80
  `Workspace mode: ${manifest.workspaceMode}`,
72
81
  ...(manifest.runtimeResolution ? [`Runtime: ${manifest.runtimeResolution.kind}`, `Runtime safety: ${manifest.runtimeResolution.safety}`, `Runtime requested: ${manifest.runtimeResolution.requestedMode}${manifest.runtimeResolution.reason ? ` (${manifest.runtimeResolution.reason})` : ""}`] : []),
73
82
  `Goal: ${manifest.goal}`,
@@ -109,5 +118,50 @@ export function handleStatus(params: TeamToolParamsValue, ctx: TeamContext): PiT
109
118
  "Recent events:",
110
119
  ...(events.length ? events.map((event) => `- ${event.time} ${event.type}${event.taskId ? ` ${event.taskId}` : ""}${event.message ? `: ${event.message}` : ""}`) : ["- (none)"]),
111
120
  ];
121
+ if (!fullDetails) {
122
+ return result(
123
+ buildCompactStatus(manifest, tasks, counts, asyncLivenessLine, phaseProgress).join("\n"),
124
+ { action: "status", status: "ok", runId: manifest.runId, artifactsRoot: manifest.artifactsRoot, intent: `status ${manifest.runId}: ${manifest.status} (compact)` },
125
+ );
126
+ }
112
127
  return result(lines.join("\n"), { action: "status", status: "ok", runId: manifest.runId, artifactsRoot: manifest.artifactsRoot, intent: `status ${manifest.runId}: ${manifest.status}` });
113
128
  }
129
+
130
+ /**
131
+ * Compact status builder (DX: Round 16 F3). A tight summary for quick checks:
132
+ * identity, status, goal, task counts, and ONLY failed / attention task
133
+ * errors — not the 40-line dense dump. Invoked when params.details === false.
134
+ *
135
+ * Exported for unit testing.
136
+ */
137
+ export function buildCompactStatus(
138
+ manifest: { runId: string; team: string; workflow?: string; status: string; goal: string; workspaceMode?: string },
139
+ tasks: Array<{ id: string; status: string; role: string; agent: string; error?: string }>,
140
+ counts: Map<string, number>,
141
+ asyncLivenessLine?: string,
142
+ progress?: { overallPercentage: number; estimatedRemainingMs: number },
143
+ ): string[] {
144
+ const failedOrAttention = tasks.filter(
145
+ (t) =>
146
+ t.status === "failed" ||
147
+ t.status === "needs_attention" ||
148
+ t.status === "cancelled",
149
+ );
150
+ const lines = [
151
+ `Run: ${manifest.runId}`,
152
+ `Team: ${manifest.team}${manifest.workflow ? ` (${manifest.workflow})` : ""}`,
153
+ `Status: ${manifest.status}`,
154
+ ...(progress ? [`Progress: ${progress.overallPercentage}% (~${formatDuration(progress.estimatedRemainingMs)} remaining)`] : []),
155
+ `Goal: ${manifest.goal}`,
156
+ ...(asyncLivenessLine ? [asyncLivenessLine] : []),
157
+ `Tasks: ${[...counts.entries()].map(([s, c]) => `${s}=${c}`).join(", ") || "none"}`,
158
+ ];
159
+ if (failedOrAttention.length > 0) {
160
+ lines.push("Issues:");
161
+ for (const t of failedOrAttention) {
162
+ lines.push(`- ${t.id} [${t.status}] ${t.role}: ${t.error ?? "(no error detail)"}`);
163
+ }
164
+ }
165
+ lines.push("Tip: pass details=true for full output (task graph, agents, effectiveness, events).");
166
+ return lines;
167
+ }
@@ -156,6 +156,8 @@ import { handleParallel } from "./team-tool/parallel-dispatch.ts";
156
156
  import { handlePlan } from "./team-tool/plan.ts";
157
157
  import { handleRespond } from "./team-tool/respond.ts";
158
158
  import { handleStatus } from "./team-tool/status.ts";
159
+ import { RUN_NOT_FOUND_HINT } from "./team-tool/run-not-found.ts";
160
+ import { formatActionSuggestion } from "./action-suggestions.ts";
159
161
 
160
162
  export { handleApi } from "./team-tool/api.ts";
161
163
  export { handleRetry } from "./team-tool/cancel.ts";
@@ -459,14 +461,14 @@ export async function handleResume(
459
461
  const runCwd = locateRunCwd(params.runId, ctx.cwd);
460
462
  if (!runCwd)
461
463
  return result(
462
- `Run '${params.runId}' not found.`,
464
+ `Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`,
463
465
  { action: "resume", status: "error" },
464
466
  true,
465
467
  );
466
468
  const loaded = loadRunManifestById(runCwd, params.runId); // NOTE: no withRunLock - best-effort only; concurrent writes may cause inconsistency
467
469
  if (!loaded)
468
470
  return result(
469
- `Run '${params.runId}' not found.`,
471
+ `Run '${params.runId}' not found.${RUN_NOT_FOUND_HINT}`,
470
472
  { action: "resume", status: "error" },
471
473
  true,
472
474
  );
@@ -1347,7 +1349,7 @@ export async function handleTeamTool(
1347
1349
  }
1348
1350
  default:
1349
1351
  return result(
1350
- `Unknown action: ${action}`,
1352
+ `Unknown action: ${action}${formatActionSuggestion(String(action))}`,
1351
1353
  { action: "unknown", status: "error" },
1352
1354
  true,
1353
1355
  );
@@ -231,6 +231,13 @@ export async function spawnBackgroundTeamRun(manifest: TeamRunManifest): Promise
231
231
  windowsHide: true,
232
232
  } as unknown as Parameters<typeof spawn>[2];
233
233
  const child = spawn(process.execPath, command.args, spawnOpts);
234
+ // Round 27 (BUG 3): the piped stdout/stderr are NEVER read or destroyed →
235
+ // 2 FDs leak per background spawn, and if the child writes >64KB (pipe
236
+ // buffer) it blocks forever (nobody drains the pipe) → background runner
237
+ // hangs. The background runner redirects its own console to a file, so we
238
+ // don't need this output — destroy the read ends immediately.
239
+ child.stdout?.destroy();
240
+ child.stderr?.destroy();
234
241
  child.on("error", (error: Error) => {
235
242
  logInternalError("async-runner.spawn", error, `pid=${child.pid ?? "unknown"}`);
236
243
  });
@@ -525,7 +525,13 @@ async function main(): Promise<void> {
525
525
  const agents = allAgents(discoverAgents(cwd));
526
526
  debugLog(`[background-runner] discoverAgents done, ${agents.length} agents`,
527
527
  );
528
- try { fs.fsyncSync(fs.openSync(manifest.eventsPath, "a")); } catch { /* best-effort */ } // FORCE flush so we see this before death
528
+ // Round 27 (BUG 2): openSync returned an fd that was never closed FD
529
+ // leak per background runner startup. Close it in a finally (matches the
530
+ // canonical pattern in checkpoint.ts:83 and event-log.ts:582).
531
+ try {
532
+ const fd = fs.openSync(manifest.eventsPath, "a");
533
+ try { fs.fsyncSync(fd); } finally { try { fs.closeSync(fd); } catch { /* best-effort */ } }
534
+ } catch { /* best-effort */ } // FORCE flush so we see this before death
529
535
  debugLog(`[background-runner] calling directTeamAndWorkflowFromRun`,
530
536
  );
531
537
  const direct = directTeamAndWorkflowFromRun(manifest, tasks, agents);