@giladbarnea/pi-simple-team 1.0.26 → 1.0.28

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.
package/README.md CHANGED
@@ -55,17 +55,45 @@ You:
55
55
  Main agent:
56
56
  > Pair created. They'll loop through it together. They'll let me know when they're done.
57
57
 
58
+ <p align="center">
59
+ <img
60
+ src="screenshots/abstract-tool-output-preview.png"
61
+ width="640"
62
+ alt="A team appears, receives direction, and starts working"
63
+ >
64
+ </p>
65
+
66
+ <div align="center"><em>Spawn, direct, and follow a team from the normal Pi conversation.</em></div>
67
+
68
+ <br>
69
+
70
+ <details>
71
+ <summary><strong>See real tool outputs</strong></summary>
72
+
58
73
  <br>
59
74
 
60
- ![Spawning a designer–adversary team, checking its status, and sending both teammates a shared instruction](screenshots/1.png)
75
+ <p align="center">
76
+ <a href="screenshots/1.png"><img src="screenshots/1.png" width="48%" alt="Pi tool outputs for spawning a team, sending instructions, and checking status"></a>
77
+ <a href="screenshots/2.png"><img src="screenshots/2.png" width="48%" alt="Pi tool outputs for team status and the timestamped event log"></a>
78
+ </p>
79
+ <p align="center">
80
+ <a href="screenshots/team-status-multiple.png"><img src="screenshots/team-status-multiple.png" width="48%" alt="Status outputs for several parallel teams"></a>
81
+ <a href="screenshots/team-log-multiple.png"><img src="screenshots/team-log-multiple.png" width="48%" alt="Event log outputs for two parallel teams"></a>
82
+ </p>
61
83
 
62
- <div align="center"><em><code>pi-simple-team</code> in the TUI.</em></div>
84
+ </details>
63
85
 
64
86
  ### Watch a live team with `/team`.
65
87
 
66
88
  `/team` is for you, not another agent tool. It gives you a live view of the room without turning the main agent into a status relay.
67
89
 
68
- The command opens an overlay rolling recent statuses, messages, and a separate event log.
90
+ The command opens an overlay rolling recent statuses, messages, and a separate event log.
91
+
92
+ <a href="screenshots/team-view.png">
93
+ <img src="screenshots/team-view.png" width="100%" alt="The live team dashboard showing statuses, messages, and recent events">
94
+ </a>
95
+
96
+ <div align="center"><em>The live, read-only <code>/team</code> dashboard.</em></div>
69
97
 
70
98
 
71
99
  ## The mechanics
@@ -121,11 +149,6 @@ You can stop babysitting everyone's context window.
121
149
 
122
150
  Your main uses `teamlog` when it needs to understand the chain of events in high granularity: a timestamped, filterable timeline of messages, tool calls, and lifecycle events.
123
151
 
124
- <br>
125
-
126
- ![A live team status map above its timestamped event log](screenshots/2.png)
127
- <div align="center"><em>Visual representation of the team log.</em></div>
128
-
129
152
  ### 🪟 Herdr Mode
130
153
 
131
154
  `pi-simple-team` can give each teammate a Pi session in its own [Herdr](https://herdr.dev) pane. Just ask your main agent to do that.
package/child-tools.ts CHANGED
@@ -7,8 +7,8 @@ import { renderTeamMessage } from "./render.ts";
7
7
  type JsonRecord = Record<string, unknown>;
8
8
 
9
9
  const teamMessageType = "pi-simple-team";
10
- const defaultVisibleInterruptWaitTimeoutMilliseconds = 25_000;
11
- const visibleLifecycleRetryAttempts = 3;
10
+ const defaultInterruptWaitTimeoutMilliseconds = 25_000;
11
+ const lifecycleCallbackRetryAttempts = 3;
12
12
 
13
13
  export interface ChildRuntimeConfig {
14
14
  callbackUrl: string;
@@ -16,7 +16,6 @@ export interface ChildRuntimeConfig {
16
16
  teamId?: string;
17
17
  teamName: string;
18
18
  teammateName: string;
19
- visible: boolean;
20
19
  participants: string[];
21
20
  canOverseeOwnTeams: boolean;
22
21
  interruptWaitTimeoutMilliseconds?: number;
@@ -45,7 +44,6 @@ function readRequiredChildRuntimeConfig(): ChildRuntimeConfig {
45
44
  teamId,
46
45
  teamName: process.env.PI_SIMPLE_TEAM_TEAM_NAME ?? teamId,
47
46
  teammateName: requiredEnvironmentVariable("PI_SIMPLE_TEAM_MEMBER"),
48
- visible: process.env.PI_SIMPLE_TEAM_VISIBLE_CHILD === "1",
49
47
  participants: readParticipants(),
50
48
  canOverseeOwnTeams: process.env.PI_SIMPLE_TEAM_CAN_OVERSEE_OWN_TEAMS === "1",
51
49
  };
@@ -99,7 +97,7 @@ async function readJsonBody(request: http.IncomingMessage): Promise<JsonRecord>
99
97
  return JSON.parse(Buffer.concat(chunks).toString("utf8")) as JsonRecord;
100
98
  }
101
99
 
102
- interface VisibleDelivery {
100
+ interface ChildDelivery {
103
101
  team: string;
104
102
  from: string;
105
103
  to: string;
@@ -109,7 +107,8 @@ interface VisibleDelivery {
109
107
  interrupt: boolean;
110
108
  }
111
109
 
112
- function startVisibleChild(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
110
+ /** The one live runtime every child gets: a delivery server, parent registration, and lifecycle callbacks. */
111
+ function startChildRuntime(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
113
112
  let server: http.Server | undefined;
114
113
  let activeContext: ExtensionContext | undefined;
115
114
  let idle = true;
@@ -120,12 +119,12 @@ function startVisibleChild(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
120
119
  const notifyParent = (event: JsonRecord): Promise<void> => {
121
120
  lifecycleQueue = lifecycleQueue.then(async () => {
122
121
  if (lifecycleError) return;
123
- for (let attempt = 1; attempt <= visibleLifecycleRetryAttempts; attempt += 1) {
122
+ for (let attempt = 1; attempt <= lifecycleCallbackRetryAttempts; attempt += 1) {
124
123
  try {
125
- await callParent(config, "visible_event", { event });
124
+ await callParent(config, "event", { event });
126
125
  return;
127
126
  } catch (error) {
128
- if (attempt === visibleLifecycleRetryAttempts) lifecycleError = error instanceof Error ? error : new Error(String(error));
127
+ if (attempt === lifecycleCallbackRetryAttempts) lifecycleError = error instanceof Error ? error : new Error(String(error));
129
128
  }
130
129
  }
131
130
  });
@@ -138,8 +137,8 @@ function startVisibleChild(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
138
137
  let resolveIdle: () => void;
139
138
  const timeout = setTimeout(() => {
140
139
  idleWaiters = idleWaiters.filter((waiter) => waiter !== resolveIdle);
141
- reject(new Error("Timed out waiting for visible child to settle after interrupt"));
142
- }, config.interruptWaitTimeoutMilliseconds ?? defaultVisibleInterruptWaitTimeoutMilliseconds);
140
+ reject(new Error("Timed out waiting for the child to settle after interrupt"));
141
+ }, config.interruptWaitTimeoutMilliseconds ?? defaultInterruptWaitTimeoutMilliseconds);
143
142
  resolveIdle = (): void => {
144
143
  clearTimeout(timeout);
145
144
  resolve();
@@ -194,8 +193,8 @@ function startVisibleChild(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
194
193
  return;
195
194
  }
196
195
 
197
- const delivery = body.args as unknown as VisibleDelivery;
198
- if (lifecycleError) throw new Error(`Visible lifecycle callback failed: ${lifecycleError.message}`);
196
+ const delivery = body.args as unknown as ChildDelivery;
197
+ if (lifecycleError) throw new Error(`Lifecycle callback failed: ${lifecycleError.message}`);
199
198
  if (delivery.interrupt && activeContext) {
200
199
  activeContext.abort();
201
200
  await waitForIdle();
@@ -228,10 +227,10 @@ function startVisibleChild(pi: ExtensionAPI, config: ChildRuntimeConfig): void {
228
227
  server!.listen(0, "127.0.0.1", () => resolve());
229
228
  });
230
229
  const address = server.address();
231
- if (!address || typeof address === "string") throw new Error("Visible teammate server did not get a port");
230
+ if (!address || typeof address === "string") throw new Error("Child delivery server did not get a port");
232
231
 
233
232
  try {
234
- await callParent(config, "visible_register", {
233
+ await callParent(config, "register", {
235
234
  url: `http://127.0.0.1:${address.port}/deliver`,
236
235
  sessionId: context.sessionManager.getSessionId(),
237
236
  sessionFile: context.sessionManager.getSessionFile(),
@@ -276,7 +275,7 @@ export function registerChildTools(pi: ExtensionAPI, config: ChildRuntimeConfig)
276
275
  ),
277
276
  };
278
277
  });
279
- if (config.visible) startVisibleChild(pi, config);
278
+ startChildRuntime(pi, config);
280
279
 
281
280
  pi.registerTool(
282
281
  defineTool({
package/index.ts CHANGED
@@ -4,7 +4,6 @@ import crypto from "node:crypto";
4
4
  import * as fs from "node:fs";
5
5
  import http from "node:http";
6
6
  import * as path from "node:path";
7
- import { StringDecoder } from "node:string_decoder";
8
7
  import { fileURLToPath } from "node:url";
9
8
  import { defineTool, type ContextUsage, type ExtensionAPI, getMarkdownTheme } from "@earendil-works/pi-coding-agent";
10
9
  import { bundledAiToAiSkillInstruction } from "./bundled-skill.ts";
@@ -40,21 +39,6 @@ interface TeamStatus {
40
39
  updated: string;
41
40
  }
42
41
 
43
- interface RpcResponse extends JsonRecord {
44
- type: "response";
45
- id?: string;
46
- command: string;
47
- success: boolean;
48
- error?: string;
49
- data?: unknown;
50
- }
51
-
52
- interface PendingRequest {
53
- resolve: (response: RpcResponse) => void;
54
- reject: (error: Error) => void;
55
- timeout: ReturnType<typeof setTimeout>;
56
- }
57
-
58
42
  interface TeammateSpec {
59
43
  name: string;
60
44
  prompt: string;
@@ -79,15 +63,12 @@ interface TeammateState {
79
63
  sessionMaterialized: boolean;
80
64
  process?: ChildProcess;
81
65
  paneId?: string;
82
- visibleUrl?: string;
83
- visibleReady?: Promise<void>;
84
- resolveVisibleReady?: () => void;
85
- rejectVisibleReady?: (error: Error) => void;
86
- busy: boolean;
66
+ deliveryUrl?: string;
67
+ ready?: Promise<void>;
68
+ resolveReady?: () => void;
69
+ rejectReady?: (error: Error) => void;
87
70
  alive: boolean;
88
- pendingRequests: Map<string, PendingRequest>;
89
71
  deliveryQueue: Promise<void>;
90
- recentEvents: JsonRecord[];
91
72
  stderr: string;
92
73
  }
93
74
 
@@ -112,14 +93,13 @@ interface TeamState {
112
93
  const teamLiteExtensionPath = fileURLToPath(import.meta.url);
113
94
  const thinkingLevels = ["low", "medium", "high", "xhigh", "max"] as const;
114
95
  const defaultThinkingLevel: ThinkingLevel = "xhigh";
115
- const visibleDeliveryTimeoutMilliseconds = 30_000;
96
+ const deliveryTimeoutMilliseconds = 30_000;
116
97
  const defaultRpcShutdownGraceMilliseconds = 1_000;
117
98
  const teamMessageType = "pi-simple-team";
118
99
  const teams = new Map<string, TeamState>();
119
100
  const callbackToken = crypto.randomBytes(24).toString("hex");
120
101
  let callbackServer: http.Server | undefined;
121
102
  let callbackUrl = "";
122
- let requestCounter = 0;
123
103
 
124
104
  function status(word: string, phrase: string): TeamStatus {
125
105
  return { word, phrase, updated: nowText() };
@@ -138,10 +118,6 @@ function toolResult(payload: JsonRecord) {
138
118
  };
139
119
  }
140
120
 
141
- function sleep(milliseconds: number): Promise<void> {
142
- return new Promise((resolve) => setTimeout(resolve, milliseconds));
143
- }
144
-
145
121
  interface CommandResult {
146
122
  stdout: string;
147
123
  stderr: string;
@@ -177,162 +153,16 @@ async function validateHerdrAvailability(): Promise<string> {
177
153
  return tabId;
178
154
  }
179
155
 
180
- function serializeJsonLine(value: unknown): string {
181
- return `${JSON.stringify(value)}\n`;
182
- }
183
-
184
- function attachJsonlReader(stream: NodeJS.ReadableStream, onLine: (line: string) => void): void {
185
- const decoder = new StringDecoder("utf8");
186
- let buffer = "";
187
-
188
- stream.on("data", (chunk: Buffer | string) => {
189
- buffer += typeof chunk === "string" ? chunk : decoder.write(chunk);
190
-
191
- while (true) {
192
- const newlineIndex = buffer.indexOf("\n");
193
- if (newlineIndex === -1) return;
194
-
195
- let line = buffer.slice(0, newlineIndex);
196
- buffer = buffer.slice(newlineIndex + 1);
197
- if (line.endsWith("\r")) line = line.slice(0, -1);
198
- onLine(line);
199
- }
200
- });
201
-
202
- stream.on("end", () => {
203
- buffer += decoder.end();
204
- if (buffer.length > 0) onLine(buffer.endsWith("\r") ? buffer.slice(0, -1) : buffer);
205
- });
206
- }
207
-
208
- function pushRecentEvent(teammate: TeammateState, event: JsonRecord): void {
209
- teammate.recentEvents.push(event);
210
- if (teammate.recentEvents.length > 200) teammate.recentEvents.shift();
211
- }
212
-
213
- function rejectPendingRequests(teammate: TeammateState, error: Error): void {
214
- for (const pending of teammate.pendingRequests.values()) {
215
- clearTimeout(pending.timeout);
216
- pending.reject(error);
217
- }
218
- teammate.pendingRequests.clear();
219
- }
220
-
221
- function handleTeammateEvent(team: TeamState, teammate: TeammateState, event: JsonRecord): void {
222
- pushRecentEvent(teammate, event);
223
-
224
- if (event.type === "agent_start") teammate.busy = true;
225
- if (event.type === "agent_end") teammate.busy = false;
226
-
227
- const logInput = normalizeChildEvent(team.name, teammate.name, event);
228
- if (logInput) appendTeamLog(team, logInput);
229
-
230
- if (event.type !== "response") return;
231
- const response = event as RpcResponse;
232
- const requestId = response.id;
233
- if (!requestId) return;
234
-
235
- const pending = teammate.pendingRequests.get(requestId);
236
- if (!pending) return;
237
-
238
- teammate.pendingRequests.delete(requestId);
239
- clearTimeout(pending.timeout);
240
- pending.resolve(response);
241
- }
242
-
243
- function sendRpc(teammate: TeammateState, command: JsonRecord, timeoutMilliseconds = 30_000): Promise<RpcResponse> {
244
- if (teammate.transport !== "rpc" || !teammate.process || !teammate.alive || teammate.process.exitCode !== null) {
245
- throw new Error(`Teammate ${teammate.name} is not alive through RPC`);
246
- }
247
-
248
- const id = `team-${++requestCounter}`;
249
- const payload = { ...command, id };
250
-
251
- return new Promise((resolve, reject) => {
252
- const timeout = setTimeout(() => {
253
- teammate.pendingRequests.delete(id);
254
- reject(new Error(`Timed out waiting for ${String(command.type)} response from ${teammate.name}`));
255
- }, timeoutMilliseconds);
256
-
257
- teammate.pendingRequests.set(id, { resolve, reject, timeout });
258
- teammate.process!.stdin?.write(serializeJsonLine(payload));
259
- });
260
- }
261
-
262
- async function requireSuccessfulResponse(response: RpcResponse, action: string): Promise<void> {
263
- if (response.success) return;
264
- throw new Error(`${action} failed: ${response.error ?? "unknown RPC error"}`);
265
- }
266
-
267
- async function getTeammateBusy(teammate: TeammateState): Promise<boolean> {
268
- if (teammate.transport === "herdr") return teammate.busy;
269
-
270
- const response = await sendRpc(teammate, { type: "get_state" }, 10_000);
271
- if (!response.success) return teammate.busy;
272
- const data = response.data as { isStreaming?: boolean } | undefined;
273
- teammate.busy = Boolean(data?.isStreaming);
274
- return teammate.busy;
275
- }
276
-
277
- async function waitForIdle(teammate: TeammateState, timeoutMilliseconds = 30_000): Promise<void> {
278
- const deadline = Date.now() + timeoutMilliseconds;
279
-
280
- while (Date.now() < deadline) {
281
- if (!(await getTeammateBusy(teammate))) return;
282
- await sleep(100);
283
- }
284
-
285
- throw new Error(`Timed out waiting for ${teammate.name} to become idle`);
286
- }
287
-
288
- async function promptTeammate(team: TeamState, teammate: TeammateState, message: string): Promise<void> {
289
- const response = await sendRpc(teammate, { type: "prompt", message });
290
- if (response.success) {
291
- appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "ack", summary: "prompt accepted" });
292
- return;
293
- }
294
-
295
- const fallback = await sendRpc(teammate, { type: "prompt", message, streamingBehavior: "steer" });
296
- if (fallback.success) {
297
- appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "ack", summary: "steer fallback accepted" });
298
- } else {
299
- appendTeamLog(team, {
300
- team: team.name,
301
- teammate: teammate.name,
302
- direction: "runtime",
303
- kind: "error",
304
- summary: preview(`steer fallback to ${teammate.name} failed: ${fallback.error ?? "unknown RPC error"}`),
305
- });
306
- }
307
- await requireSuccessfulResponse(fallback, `steer fallback to ${teammate.name}`);
308
- }
309
-
310
- async function steerTeammate(team: TeamState, teammate: TeammateState, message: string): Promise<void> {
311
- const response = await sendRpc(teammate, { type: "steer", message });
312
- if (response.success) {
313
- appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "ack", summary: "steer accepted" });
314
- return;
315
- }
316
- appendTeamLog(team, {
317
- team: team.name,
318
- teammate: teammate.name,
319
- direction: "runtime",
320
- kind: "error",
321
- summary: preview(`steer to ${teammate.name} rejected: ${response.error ?? "unknown RPC error"}`),
322
- });
323
- await promptTeammate(team, teammate, message);
324
- }
325
-
326
156
  function formatTeammateMessage(team: TeamState, from: string, message: string): string {
327
157
  return [`[from ${from} on team ${team.name}]`, message, "", "Current team status:", JSON.stringify(formatStatus(team), null, 2)].join("\n");
328
158
  }
329
159
 
330
- async function deliverVisibleMessage(team: TeamState, from: string, recipient: TeammateState, message: string, formattedMessage: string, interrupt: boolean): Promise<void> {
331
- if (!recipient.alive || !recipient.visibleUrl) throw new Error(`Visible teammate ${recipient.name} is not ready`);
160
+ async function deliverMessage(team: TeamState, from: string, recipient: TeammateState, message: string, formattedMessage: string, interrupt: boolean): Promise<void> {
161
+ if (!recipient.alive || !recipient.deliveryUrl) throw new Error(`Teammate ${recipient.name} is not ready`);
332
162
  const controller = new AbortController();
333
- const timeout = setTimeout(() => controller.abort(), visibleDeliveryTimeoutMilliseconds);
163
+ const timeout = setTimeout(() => controller.abort(), deliveryTimeoutMilliseconds);
334
164
  try {
335
- const response = await fetch(recipient.visibleUrl, {
165
+ const response = await fetch(recipient.deliveryUrl, {
336
166
  method: "POST",
337
167
  headers: { "content-type": "application/json" },
338
168
  body: JSON.stringify({
@@ -342,12 +172,12 @@ async function deliverVisibleMessage(team: TeamState, from: string, recipient: T
342
172
  }),
343
173
  signal: controller.signal,
344
174
  });
345
- if (!response.ok) throw new Error(`Visible teammate ${recipient.name} rejected delivery: ${response.status} ${await response.text()}`);
175
+ if (!response.ok) throw new Error(`Teammate ${recipient.name} rejected delivery: ${response.status} ${await response.text()}`);
346
176
  const result = (await response.json()) as { accepted?: boolean };
347
- if (!result.accepted) throw new Error(`Visible teammate ${recipient.name} did not accept delivery`);
348
- appendTeamLog(team, { team: team.name, teammate: recipient.name, direction: "runtime", kind: "ack", summary: "visible message accepted" });
177
+ if (!result.accepted) throw new Error(`Teammate ${recipient.name} did not accept delivery`);
178
+ appendTeamLog(team, { team: team.name, teammate: recipient.name, direction: "runtime", kind: "ack", summary: "message accepted" });
349
179
  } catch (error) {
350
- if (controller.signal.aborted) throw new Error(`Timed out waiting for visible teammate ${recipient.name} delivery`);
180
+ if (controller.signal.aborted) throw new Error(`Timed out waiting for teammate ${recipient.name} delivery`);
351
181
  throw error;
352
182
  } finally {
353
183
  clearTimeout(timeout);
@@ -364,24 +194,7 @@ async function deliverToTeammate(team: TeamState, from: string, recipient: Teamm
364
194
  summary: preview(message),
365
195
  details: { from, to: recipient.name, interrupt },
366
196
  });
367
- if (recipient.transport === "herdr") {
368
- await deliverVisibleMessage(team, from, recipient, message, formattedMessage, interrupt);
369
- return;
370
- }
371
-
372
- if (interrupt && (recipient.busy || (await getTeammateBusy(recipient)))) {
373
- await requireSuccessfulResponse(await sendRpc(recipient, { type: "abort" }, 60_000), `abort ${recipient.name}`);
374
- await waitForIdle(recipient);
375
- await promptTeammate(team, recipient, formattedMessage);
376
- return;
377
- }
378
-
379
- if (recipient.busy || (await getTeammateBusy(recipient))) {
380
- await steerTeammate(team, recipient, formattedMessage);
381
- return;
382
- }
383
-
384
- await promptTeammate(team, recipient, formattedMessage);
197
+ await deliverMessage(team, from, recipient, message, formattedMessage, interrupt);
385
198
  }
386
199
 
387
200
  function enqueueDelivery(team: TeamState, from: string, recipient: TeammateState, message: string, interrupt: boolean): void {
@@ -454,21 +267,14 @@ function resolveContextTargets(owner: symbol, targetNames: string[]): TeammateSt
454
267
  }
455
268
 
456
269
  async function getTeammateContextUsage(teammate: TeammateState, signal?: AbortSignal): Promise<KnownContextUsage> {
457
- if (teammate.transport === "rpc") {
458
- const response = await sendRpc(teammate, { type: "get_session_stats" });
459
- await requireSuccessfulResponse(response, `read context usage from ${teammate.name}`);
460
- const stats = response.data as { contextUsage?: ContextUsage };
461
- return requireKnownContextUsage(stats.contextUsage);
462
- }
463
-
464
- if (!teammate.alive || !teammate.visibleUrl) throw new Error(`Visible teammate ${teammate.name} is not ready`);
465
- const response = await fetch(teammate.visibleUrl, {
270
+ if (!teammate.alive || !teammate.deliveryUrl) throw new Error(`Teammate ${teammate.name} is not ready`);
271
+ const response = await fetch(teammate.deliveryUrl, {
466
272
  method: "POST",
467
273
  headers: { "content-type": "application/json" },
468
274
  body: JSON.stringify({ token: callbackToken, tool: "report_context_window", args: {} }),
469
275
  signal,
470
276
  });
471
- if (!response.ok) throw new Error(`Visible teammate ${teammate.name} rejected context-window query: ${response.status} ${await response.text()}`);
277
+ if (!response.ok) throw new Error(`Teammate ${teammate.name} rejected context-window query: ${response.status} ${await response.text()}`);
472
278
  const payload = (await response.json()) as { contextUsage?: ContextUsage };
473
279
  return requireKnownContextUsage(payload.contextUsage);
474
280
  }
@@ -516,14 +322,12 @@ function logStatusDeclaration(team: TeamState, participant: string, word?: strin
516
322
  function createTeammateState(team: TeamState, teammateSpec: TeammateSpec): TeammateState {
517
323
  const teammateName = compactName(teammateSpec.name);
518
324
  const thinking = teammateSpec.thinking ?? defaultThinkingLevel;
519
- let resolveVisibleReady: (() => void) | undefined;
520
- let rejectVisibleReady: ((error: Error) => void) | undefined;
521
- const visibleReady = team.showOnHerdrPanes
522
- ? new Promise<void>((resolve, reject) => {
523
- resolveVisibleReady = resolve;
524
- rejectVisibleReady = reject;
525
- })
526
- : undefined;
325
+ let resolveReady: (() => void) | undefined;
326
+ let rejectReady: ((error: Error) => void) | undefined;
327
+ const ready = new Promise<void>((resolve, reject) => {
328
+ resolveReady = resolve;
329
+ rejectReady = reject;
330
+ });
527
331
 
528
332
  return {
529
333
  name: teammateName,
@@ -534,22 +338,18 @@ function createTeammateState(team: TeamState, teammateSpec: TeammateSpec): Teamm
534
338
  canOverseeOwnTeams: Boolean(teammateSpec.canOverseeOwnTeams),
535
339
  transport: team.showOnHerdrPanes ? "herdr" : "rpc",
536
340
  sessionMaterialized: false,
537
- visibleReady,
538
- resolveVisibleReady,
539
- rejectVisibleReady,
540
- busy: false,
341
+ ready,
342
+ resolveReady,
343
+ rejectReady,
541
344
  alive: true,
542
- pendingRequests: new Map(),
543
345
  deliveryQueue: Promise.resolve(),
544
- recentEvents: [],
545
346
  stderr: "",
546
347
  };
547
348
  }
548
349
 
549
- function childEnvironmentOverrides(team: TeamState, teammate: TeammateState, participants: string[], visible: boolean): Record<string, string> {
350
+ function childEnvironmentOverrides(team: TeamState, teammate: TeammateState, participants: string[]): Record<string, string> {
550
351
  return {
551
352
  PI_SIMPLE_TEAM_CHILD: "1",
552
- PI_SIMPLE_TEAM_VISIBLE_CHILD: visible ? "1" : "0",
553
353
  PI_SIMPLE_TEAM_CALLBACK_URL: callbackUrl,
554
354
  PI_SIMPLE_TEAM_CALLBACK_TOKEN: callbackToken,
555
355
  PI_SIMPLE_TEAM_TEAM: team.id ?? team.name,
@@ -571,12 +371,12 @@ function appendSpawnLog(team: TeamState, teammate: TeammateState): void {
571
371
  });
572
372
  }
573
373
 
574
- interface RpcStartOptions {
374
+ interface ChildStartOptions {
575
375
  sessionFile?: string;
576
376
  restartEmpty?: boolean;
577
377
  }
578
378
 
579
- function attachRpcTeammate(team: TeamState, teammate: TeammateState, participants: string[], options: RpcStartOptions): void {
379
+ function attachRpcTeammate(team: TeamState, teammate: TeammateState, participants: string[], options: ChildStartOptions): void {
580
380
  const sessionArgs = options.sessionFile
581
381
  ? ["--session", options.sessionFile]
582
382
  : teammate.inheritContext && !options.restartEmpty
@@ -598,21 +398,12 @@ function attachRpcTeammate(team: TeamState, teammate: TeammateState, participant
598
398
  ];
599
399
  const proc = childProcess.spawn("pi", args, {
600
400
  cwd: team.projectDirectory ?? process.cwd(),
601
- stdio: ["pipe", "pipe", "pipe"],
602
- env: { ...process.env, ...childEnvironmentOverrides(team, teammate, participants, false) },
401
+ stdio: ["pipe", "ignore", "pipe"],
402
+ env: { ...process.env, ...childEnvironmentOverrides(team, teammate, participants) },
603
403
  });
604
404
  teammate.process = proc;
605
405
  teammate.alive = true;
606
- teammate.busy = false;
607
406
 
608
- attachJsonlReader(proc.stdout!, (line) => {
609
- if (!line.trim()) return;
610
- try {
611
- handleTeammateEvent(team, teammate, JSON.parse(line) as JsonRecord);
612
- } catch {
613
- teammate.stderr += `\n[unparsed stdout] ${line}`;
614
- }
615
- });
616
407
  proc.stderr?.on("data", (chunk: Buffer | string) => {
617
408
  const text = typeof chunk === "string" ? chunk : chunk.toString("utf8");
618
409
  teammate.stderr += text;
@@ -620,26 +411,13 @@ function attachRpcTeammate(team: TeamState, teammate: TeammateState, participant
620
411
  });
621
412
  proc.on("exit", (code, signal) => {
622
413
  teammate.alive = false;
623
- teammate.busy = false;
624
- rejectPendingRequests(teammate, new Error(`${teammate.name} exited (code=${code}, signal=${signal})`));
414
+ teammate.rejectReady?.(new Error(`${teammate.name} exited (code=${code}, signal=${signal})`));
625
415
  team.statuses.set(teammate.name, status("stopped", `Exited code=${code} signal=${signal}`));
626
416
  appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "exit", summary: `exited (code=${code}, signal=${signal})`, details: { code, signal } });
627
417
  });
628
418
  appendSpawnLog(team, teammate);
629
419
  }
630
420
 
631
- async function captureRpcSessionIdentity(teammate: TeammateState): Promise<void> {
632
- const response = await sendRpc(teammate, { type: "get_state" }, 10_000);
633
- await requireSuccessfulResponse(response, `read session identity from ${teammate.name}`);
634
- const state = response.data as { sessionId?: unknown; sessionFile?: unknown } | undefined;
635
- if (typeof state?.sessionId !== "string" || typeof state.sessionFile !== "string" || !path.isAbsolute(state.sessionFile)) {
636
- throw new Error(`Teammate ${teammate.name} reported an invalid session identity`);
637
- }
638
- teammate.sessionId = state.sessionId;
639
- teammate.sessionFile = state.sessionFile;
640
- teammate.sessionMaterialized = fs.existsSync(state.sessionFile);
641
- }
642
-
643
421
  function parseHerdrPaneId(stdout: string, teammateName: string): string {
644
422
  const response = JSON.parse(stdout) as { result?: { agent?: { pane_id?: string } } };
645
423
  const paneId = response.result?.agent?.pane_id;
@@ -647,15 +425,15 @@ function parseHerdrPaneId(stdout: string, teammateName: string): string {
647
425
  return paneId;
648
426
  }
649
427
 
650
- async function attachVisibleTeammate(
428
+ async function attachHerdrTeammate(
651
429
  team: TeamState,
652
430
  teammate: TeammateState,
653
431
  participants: string[],
654
432
  herdrTabId: string,
655
- options: RpcStartOptions,
433
+ options: ChildStartOptions,
656
434
  ): Promise<void> {
657
435
  const systemPrompt = composeSystemPrompt(team.name, team.teamPrompt, teammate.name, teammate.prompt, participants, teammate.canOverseeOwnTeams);
658
- const environment = childEnvironmentOverrides(team, teammate, participants, true);
436
+ const environment = childEnvironmentOverrides(team, teammate, participants);
659
437
  const sessionArgs = options.sessionFile
660
438
  ? ["--session", options.sessionFile]
661
439
  : teammate.inheritContext && !options.restartEmpty
@@ -680,17 +458,20 @@ async function attachVisibleTeammate(
680
458
  const result = await runCommand("herdr", args);
681
459
  teammate.paneId = parseHerdrPaneId(result.stdout, teammate.name);
682
460
  appendSpawnLog(team, teammate);
461
+ }
683
462
 
463
+ /** Every child registers its delivery runtime through the parent callback; startup completes only after registration. */
464
+ async function awaitChildRegistration(teammate: TeammateState): Promise<void> {
684
465
  let readinessTimeout: ReturnType<typeof setTimeout> | undefined;
685
466
  try {
686
467
  await Promise.race([
687
- teammate.visibleReady!,
468
+ teammate.ready!,
688
469
  new Promise<never>((_, reject) => {
689
- readinessTimeout = setTimeout(() => reject(new Error(`Timed out waiting for visible teammate ${teammate.name} readiness`)), 30_000);
470
+ readinessTimeout = setTimeout(() => reject(new Error(`Timed out waiting for teammate ${teammate.name} to register`)), 30_000);
690
471
  }),
691
472
  ]);
692
473
  } catch (error) {
693
- teammate.rejectVisibleReady?.(error instanceof Error ? error : new Error(String(error)));
474
+ teammate.rejectReady?.(error instanceof Error ? error : new Error(String(error)));
694
475
  throw error;
695
476
  } finally {
696
477
  if (readinessTimeout) clearTimeout(readinessTimeout);
@@ -702,14 +483,11 @@ async function startTeammate(
702
483
  teammate: TeammateState,
703
484
  participants: string[],
704
485
  herdrTabId?: string,
705
- rpcOptions: RpcStartOptions = {},
486
+ startOptions: ChildStartOptions = {},
706
487
  ): Promise<void> {
707
- if (teammate.transport === "herdr") {
708
- await attachVisibleTeammate(team, teammate, participants, herdrTabId!, rpcOptions);
709
- return;
710
- }
711
- attachRpcTeammate(team, teammate, participants, rpcOptions);
712
- if (team.id) await captureRpcSessionIdentity(teammate);
488
+ if (teammate.transport === "herdr") await attachHerdrTeammate(team, teammate, participants, herdrTabId!, startOptions);
489
+ else attachRpcTeammate(team, teammate, participants, startOptions);
490
+ await awaitChildRegistration(teammate);
713
491
  }
714
492
 
715
493
  function manifestMemberFromTeammate(teammate: TeammateState): TeamManifestMember {
@@ -750,7 +528,7 @@ function isHerdrPaneNotFound(error: unknown): boolean {
750
528
  return error instanceof Error && error.message.includes('"code":"pane_not_found"');
751
529
  }
752
530
 
753
- async function closeVisiblePane(teammate: TeammateState): Promise<void> {
531
+ async function closeHerdrPane(teammate: TeammateState): Promise<void> {
754
532
  if (!teammate.paneId) return;
755
533
  try {
756
534
  await runCommand("herdr", ["pane", "close", teammate.paneId], 10_000);
@@ -786,7 +564,7 @@ async function shutdownTeam(team: TeamState): Promise<string[]> {
786
564
  if (teammate.transport === "herdr") {
787
565
  teammate.alive = false;
788
566
  try {
789
- await closeVisiblePane(teammate);
567
+ await closeHerdrPane(teammate);
790
568
  } catch (error) {
791
569
  errors.push(error instanceof Error ? error.message : String(error));
792
570
  }
@@ -861,38 +639,29 @@ function writeJson(response: http.ServerResponse, statusCode: number, payload: J
861
639
  response.end(body);
862
640
  }
863
641
 
864
- function validateVisibleChildUrl(rawUrl: string, teammateName: string): string {
642
+ function validateChildDeliveryUrl(rawUrl: string, teammateName: string): string {
865
643
  let url: URL;
866
644
  try {
867
645
  url = new URL(rawUrl);
868
646
  } catch {
869
- throw new Error(`Invalid visible teammate URL for ${teammateName}`);
647
+ throw new Error(`Invalid delivery URL for ${teammateName}`);
870
648
  }
871
649
  const port = Number(url.port);
872
650
  if (url.protocol !== "http:" || url.hostname !== "127.0.0.1" || !url.port || !Number.isInteger(port) || port < 1 || port > 65_535 || url.pathname !== "/deliver") {
873
- throw new Error(`Invalid visible teammate URL for ${teammateName}`);
651
+ throw new Error(`Invalid delivery URL for ${teammateName}`);
874
652
  }
875
653
  return url.toString();
876
654
  }
877
655
 
878
- function handleVisibleEvent(team: TeamState, teammate: TeammateState, event: JsonRecord): void {
879
- if (event.type === "visible_startup_error") {
880
- const error = new Error(String(event.error ?? "Visible teammate startup failed"));
881
- teammate.rejectVisibleReady?.(error);
882
- throw error;
883
- }
884
- if (event.type === "agent_start") teammate.busy = true;
885
- if (event.type === "agent_end") teammate.busy = false;
656
+ function handleChildEvent(team: TeamState, teammate: TeammateState, event: JsonRecord): void {
886
657
  if (event.type === "session_shutdown") {
887
658
  teammate.alive = false;
888
- teammate.busy = false;
889
- teammate.visibleUrl = undefined;
659
+ teammate.deliveryUrl = undefined;
890
660
  if (event.reason === "quit") {
891
661
  team.statuses.set(teammate.name, status("stopped", "Session shut down"));
892
- appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "exit", summary: "visible session shut down", details: { reason: event.reason } });
662
+ appendTeamLog(team, { team: team.name, teammate: teammate.name, direction: "runtime", kind: "exit", summary: "session shut down", details: { reason: event.reason } });
893
663
  }
894
664
  }
895
- pushRecentEvent(teammate, event);
896
665
  const logInput = normalizeChildEvent(team.name, teammate.name, event);
897
666
  if (logInput) appendTeamLog(team, logInput);
898
667
  }
@@ -911,35 +680,34 @@ async function handleCallbackRequest(request: http.IncomingMessage, response: ht
911
680
  const args = (body.args ?? {}) as JsonRecord;
912
681
  const teammate = team.members.get(from);
913
682
 
914
- if (tool === "visible_register") {
915
- if (!teammate || teammate.transport !== "herdr") throw new Error(`Unknown visible teammate: ${from}`);
683
+ if (tool === "register") {
684
+ if (!teammate) throw new Error(`Unknown teammate: ${from}`);
916
685
  const url = String(args.url ?? "");
917
686
  const sessionId = args.sessionId;
918
687
  const sessionFile = args.sessionFile;
919
688
  if (typeof sessionId !== "string" || typeof sessionFile !== "string" || !path.isAbsolute(sessionFile)) {
920
- throw new Error(`Visible teammate ${from} reported an invalid session identity`);
689
+ throw new Error(`Teammate ${from} reported an invalid session identity`);
921
690
  }
922
691
  try {
923
- teammate.visibleUrl = validateVisibleChildUrl(url, from);
692
+ teammate.deliveryUrl = validateChildDeliveryUrl(url, from);
924
693
  } catch (error) {
925
694
  const failure = error instanceof Error ? error : new Error(String(error));
926
- teammate.rejectVisibleReady?.(failure);
695
+ teammate.rejectReady?.(failure);
927
696
  throw failure;
928
697
  }
929
698
  teammate.sessionId = sessionId;
930
699
  teammate.sessionFile = sessionFile;
931
700
  teammate.sessionMaterialized = fs.existsSync(sessionFile);
932
701
  teammate.alive = true;
933
- teammate.busy = false;
934
702
  team.statuses.set(teammate.name, status("idle", "Spawned"));
935
- teammate.resolveVisibleReady?.();
703
+ teammate.resolveReady?.();
936
704
  writeJson(response, 200, { accepted: true, team: team.name, from });
937
705
  return;
938
706
  }
939
707
 
940
- if (tool === "visible_event") {
941
- if (!teammate || teammate.transport !== "herdr") throw new Error(`Unknown visible teammate: ${from}`);
942
- handleVisibleEvent(team, teammate, (args.event ?? {}) as JsonRecord);
708
+ if (tool === "event") {
709
+ if (!teammate) throw new Error(`Unknown teammate: ${from}`);
710
+ handleChildEvent(team, teammate, (args.event ?? {}) as JsonRecord);
943
711
  writeJson(response, 200, { accepted: true, team: team.name, from });
944
712
  return;
945
713
  }
@@ -1046,11 +814,11 @@ function restoreTeamState(owner: symbol, ownerPi: ExtensionAPI, manifest: TeamMa
1046
814
  return team;
1047
815
  }
1048
816
 
1049
- function prepareVisibleTeammate(teammate: TeammateState): void {
1050
- teammate.transport = "herdr";
1051
- teammate.visibleReady = new Promise<void>((resolve, reject) => {
1052
- teammate.resolveVisibleReady = resolve;
1053
- teammate.rejectVisibleReady = reject;
817
+ function prepareTeammateStart(teammate: TeammateState, transport: TeammateTransport): void {
818
+ teammate.transport = transport;
819
+ teammate.ready = new Promise<void>((resolve, reject) => {
820
+ teammate.resolveReady = resolve;
821
+ teammate.rejectReady = reject;
1054
822
  });
1055
823
  }
1056
824
 
@@ -1320,8 +1088,7 @@ export default function (pi: ExtensionAPI) {
1320
1088
 
1321
1089
  try {
1322
1090
  for (const { teammate, sessionFile } of starts) {
1323
- if (showOnHerdrPanes) prepareVisibleTeammate(teammate);
1324
- else teammate.transport = "rpc";
1091
+ prepareTeammateStart(teammate, showOnHerdrPanes ? "herdr" : "rpc");
1325
1092
  await startTeammate(team, teammate, participantNames, herdrTabId, {
1326
1093
  sessionFile,
1327
1094
  restartEmpty: sessionFile === undefined,
@@ -1338,7 +1105,7 @@ export default function (pi: ExtensionAPI) {
1338
1105
  for (const { teammate } of starts) {
1339
1106
  if (teammate.transport === "herdr") {
1340
1107
  teammate.alive = false;
1341
- await closeVisiblePane(teammate);
1108
+ await closeHerdrPane(teammate);
1342
1109
  } else {
1343
1110
  await stopRpcTeammate(teammate);
1344
1111
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@giladbarnea/pi-simple-team",
3
- "version": "1.0.26",
3
+ "version": "1.0.28",
4
4
  "description": "A flat, live team of Pi agents.",
5
5
  "keywords": [
6
6
  "pi-package",
@@ -26,8 +26,12 @@
26
26
  "skills",
27
27
  "screenshots/1.png",
28
28
  "screenshots/2.png",
29
- "screenshots/team-graph.png",
29
+ "screenshots/abstract-tool-output-preview.png",
30
30
  "screenshots/pi-simple-team-banner.png",
31
+ "screenshots/team-graph.png",
32
+ "screenshots/team-log-multiple.png",
33
+ "screenshots/team-status-multiple.png",
34
+ "screenshots/team-view.png",
31
35
  "README.md",
32
36
  "LICENSE"
33
37
  ],
Binary file
@@ -5,86 +5,78 @@ description: Best practices for getting a step-function leap in performance from
5
5
 
6
6
  ## Why delegate at all? Rationale
7
7
 
8
- The main agent already holds all the context, so why not have it just do the work directly? The answer is **context management**. Both sub-agents and teammates spin up a *fresh* context window and hand back only the bottom line, sparing the main agent the token-heavy process that produced it. Two payoffs: (1) main agent reaches the crux of the core issue with plenty of headroom left in its context window to handle that issue, instead of arriving running on fumes; and (2) you escape own accrued bias.
8
+ <!-- todo: this section could just as well be called "Why it's good to let others do work too", and framed not as main-targeted but as a rank-agnostic collaborative mindset that gets more, better work done. -->
9
+ The main agent already holds all the context, so why not have it just do the work directly? The answer is **context management**. Sub-agents and teammates own a separate context window. They pay tokens with their context while main’s stays unused. This gives main a longer runway.
10
+ Delegates hand back only the bottom line (the end product), sparing the main agent the token-heavy process that produced it.
11
+ Two payoffs:
12
+ 1. main agent reaches the crux of the core issue with plenty of headroom left in its context window to handle that issue, instead of arriving running on fumes (or worse, running out prematurely); and
13
+ 2. main agent escapes own accrued bias.
9
14
 
10
15
  The flip side: if none of these payoffs apply, don't delegate. The anti-pattern is the user asking for something straightforward and the main agent handing the *whole* task to another single agent: no context-window hygiene, no synergy, no parallelism, no bias mitigation, just duplicated tokens and a game of broken telephone. Redundant middle-management.
11
16
 
12
17
  **Abstract delegation use cases:**
13
- - **Dirty work**: you need the side effect of some process, not the process itself. Searches, summaries, reviews, implementation of a well-defined spec. Can potentially be fanned-out in parallel.
18
+ - **Dirty work**: main agent needs the side effect of some process, not the process itself. Searches, summaries, reviews, implementation of a well-defined spec. Moreoever, a few of these can potentially be fanned-out in parallel.
14
19
  - **Hivemind:** teams communicating internally in real time, as each teammate does work. Hollywood analogy: elite soldier squad or a team of spies, global channel earpieces, deployed behind enemy lines, each progresses their part, each part essential for completing the mission, global channel earpieces, consistent live updates, uncertainties surfaced to get advice, discussions to make best decisions for mission, path forward of an individual adapts in real time due to new peer finding, HQ possibly listening, warning and steering as required.
15
- - **De-bias:** It’s easier to spot blind spots or flaws when looking at someone else’s work (review; live or after the fact). Alternatively, sometimes a hypothesis is considered true only after it is confirmed independently by multiple actors (consensus).
20
+ - **De-bias:** It’s easier to spot blind spots or flaws when looking at someone else’s work (review; live or after the fact). Related: sometimes a hypothesis is considered true only after it is confirmed independently by multiple actors (consensus).
16
21
 
17
- Read [`references/firstmate.md`](references/firstmate.md) when the user asks you to serve as their First Mate. Read [`references/captain.md`](references/captain.md) when the user asks you to serve as Captain of a fleet of teams. See `./references/` for specific delegation shapes.
22
+ Read [`references/firstmate.md`](references/firstmate.md) when the user asks you to serve as their First Mate. Read [`references/captain.md`](references/captain.md) when the user asks you to serve as Captain of a fleet of teams. `./references/` contains additional resources for specific delegation use-cases and shapes.
18
23
 
19
24
  ---
20
25
 
21
- ## How to delegate work to other AI agents
22
-
23
- > Note: this section applies only to AI’s that do not mechanically inherit the session’s context window. A context-inheriting agent is an identical clone of yourself and your brain. Do not give it context because it already knows 100% of what you know. Just tell it what to go for.
24
-
25
- 1. Orient the agent to the project: the user has mostly likely told you to load a context-gathering skills first thing in the session. Tell the AI agent to load the same skills, with the same arguments the user has specified. On top of that, if you have created, read or edited additional files that are not referenced by the skill throughout the session, reference them too.
26
-
27
- 2. Be generous in giving the agent wider context—understanding *why* it’s performing the task will boost its performance. Don't micromanage or over-instruct it. The agent already has the same system prompt as you do out of the box (e.g. global and project-scoped `CLAUDE.md` or `AGENTS.md`). It is essentially an equivalent instantiation of yourself. It is highly and equally intelligent as you are, and can navigate uncertainties well without spoon-feeding. Think: What kind of input do YOU thrive on? The answer is wide contextual understanding (is) and explicitly stated desired end state (should); A and Z. Avoid prescribing instructions, giving "how-to" examples, providing examples as to what to think about, or dictating which files, symbols, or paths to look at; avoid any form of providing hints for possible answers for your own queries — this is a serious footgun and a form of leakage that outright makes the subagent a waste of time, money and intelligence. Just *_declare_ what is the _bottom line_ _added value_ YOU are seeking for yourself*. Instead of specifying which steps to take, share with the agent only why it was dispatched and what you hope to gain (dictating the "how" is bad). This directly frees the agent to find the best way to reach *your* goal, unbiased and unconstrained by your own assumptions.
28
- Essentially, all the “Don’ts” above over-fit the agent.
29
- <negative-example-1 why-bad="main agent shoots its own foot by limiting the sub-agent’s research scope">
30
- User to main agent: "Why does Vercel claim their integrated version is beneficial?"
31
- Main agent spawns a sub-agent and prompts it: "Research why Vercel claim their integrated version is beneficial (edge runtime, seamless DX, zero-config, billing, monitoring, tight coupling to `vercel` CLI / dashboard / functions)."
32
- </negative-example-1>
33
- <positive-example-1 why-good="main agent declares the bottom line added value it needs without prescribing what and how to do it">
34
- User to main agent: "Why does Vercel claim their integrated version is beneficial?"
35
- Main agent spawns a sub-agent and prompts it: "I want to know why Vercel claim their integrated version is beneficial."
36
- </positive-example-1>
37
-
38
- Example 2 settings: the `load-context` skill instructs to read CLAUDE.md, ARCHITECTURE.md, docs/webserver/API.md, docs/data/architecture.md, server/api.py, and server/db.py.
39
- <negative-example-2 why-bad="main agent fails to leverage the harness and instead prescribes what to do; moreover it makes the same scope-narrowing mistake as in example-1">
40
- User to main agent: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer. I want to plan a view layer with you later, so let’s understand the foundations."
41
- Main agent spawns a sub-agent and prompts it: "Read CLAUDE.md, ARCHITECTURE.md, docs/webserver/API.md, docs/data/architecture.md, server/api.py, server/db.py, and summarize how the REST API and data layers work. Cover how function `server/api.py:from_db` fetches the data by calling the `server/db.py:get_data` function, and how [...proceeds to prescribe ironically specific locations to “discover”]"
42
- </negative-example-2>
43
- <positive-example-2 why-good="main agent recognizes the work can be distributed concurrently, shortly shares the wider context (the “why”), forwards the user’s context levers — keeping the shared domain and handing each agent one of the two subdomains rather than dropping them — and does not micro-manage the agents with how-exactly instructions">
44
- User to main agent: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer. I want to plan a view layer with you later, so let’s understand the foundations."
45
- Main agent fans out research scope horizontally to two parallel sub-agents and prompts them: "The user and I are planning a new view layer, so we need a thorough understanding of the foundations. [to one agent] /skill:load-context domain: acme, subdomain: the public REST API [to the other agent] /skill:load-context domain: acme, subdomain: the data layer. [to both] Study your subdomain deeply and exhaustively."
46
- [Main agent receives the two independent sub-agents’ responses, thinks hard to synthesize them]
47
- Main agent responds to user: "I have deep understanding of both layers and their relationships. What did you have in mind?"
48
- </positive-example-2>
49
-
50
-
51
- <positive-example-3 why-good="main agent writes as simple a prompt as possible. adds only what is necessary to get the reviewer up to speed. does not put words in the user’s mouth." note="this example has no negative counterpart. this is a positive example.">
52
- User to main agent: "Spawn a sub-agent to review your work. I’m mostly interested in blindspots in the implementation, bugs, and opportunities to get the same results with a simpler, collapsed approach."
53
- Main agent spawns a sub-agent and prompts it: "load the load-project-context and peer-review skills. the user asked me to <user request>. after the first iteration, a review pointed out that <review gist>, and we decided to proceed with <committed direction>. i’ve implemented it. review my work (the current main dirty tree). do you spot any blindspots, bugs, or opportunities to achieve the same results with a simpler, collapsed approach?"
54
- </positive-example-3>
55
-
56
- ### Subagents and Teams are two different things
57
-
58
- 3. Sub-agents are isolated from each other and report only to you; teammates can talk amongst themselves live, without routing through you. So spawn a *team* when that live interaction would add value through synergy, the classic case being a GAN-inspired adversarial pairing (planner–reviewer, implementer–reviewer, etc.) where one produces and the other pokes holes until both are content. Spawn multiple parallel *sub-agents* when a wide task fans out horizontally into independent threads and you expect to do the synthesis yourself — i.e. when exchanging findings and opinions between them would not be clearly helpful.
59
-
60
- 4. Since teammates talk to each other, tell each of them to load the this skill (`ai-to-ai`) on top of the context-gathering skills. If you are spawning an adversary among them, tell it to load the `peer-review` skill too.
61
-
62
- Example 4 settings: at the session’s start the user ran `/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer`; the main session explored the code, and the user approved a plan to add rate limiting to the public REST API.
63
- <negative-example-4 why-bad="main agent burns its own context shuttling the diff and the feedback back and forth — dives into the sub-agent’s work and clogs its own context window worse than doing the task solo would have, acts as a reviewer when biased">
64
- User to main agent: "Great, go ahead and build it."
65
- Main agent spawns one sub-agent to implement; when it returns the diff, studies and reviews it; relays the review the sub-agent; and keeps ferrying revisions until the diff settles.
66
- </negative-example-4>
67
- <positive-example-4 why-good="main agent picks a team because the adversarial iteration is synergistic, replicates the user’s context levers verbatim — including the domain and the subdomains the user specified when loading the context skill — has the reviewer also load `peer-review`, declares only the bottom line it wants, and stays out of the loop while they converge">
68
- User to main agent: "Great, go ahead and implement the plan."
69
- Main agent spawns an implementer–reviewer team and prompts them: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer, then load `ai-to-ai`. You are an implementer–reviewer team. Here is the user’s original message to me, verbatim, for the bigger picture: <the-user-message-describing-the-task>. [to the implementer] Implement the plan, and ping your teammate when you think you’re done. [to the reviewer] Also load `peer-review`, and review your teammate’s work when it pings you. [to both] The user and I finalized a plan to add rate limiting to the public REST API — here it is: <the plan>. Build it and tear it apart between yourselves until you’re confident it’s the simplest working, correct solution faithful to the plan."
70
- [The team implements and reviews live, converging without the main agent in the loop; the main agent receives the finished, reviewed result.]
71
- Main agent responds to user: "Done — implemented and adversarially reviewed between the two of them. Here’s what landed: …"
72
- </positive-example-4>
73
-
74
- 5. Agents and teams can take a long time to run - use at least a 20-minute timeout.
75
-
76
- 6. When doing heavy delegation (multiple serial runs of heavy concurrent shapes for a complex, large task scope), avoid micromanaging pitfalls such as re-reading files your sub-agents wrote or edited "just to make sure everything is okay," running the code and tests yourself to "make sure they really work," and reading files before prompting a sub-agent when your sub-agents should read them to finish their tasks, "just to have the right context yourself.". Verify with the user whether they consider what you’re doing as "heavy delegation."
26
+ ## How to request work from other AI agents (that do not inherit your session’s context window)
27
+
28
+ > Note: this section applies only to AI’s that do not mechanically inherit your own session’s context window. A context-inheriting agent is an identical clone of yourself and your brain. Do not give it context because it already knows 100% of what you know. Just tell it "You are the fork; do X". Ten words.
29
+
30
+ This section is a communication rulebook to all cases where an AI requests something from another AI: down, sideways and up — main to sub-agent, main to team, teammate to teammate, and delegated to delegator.
31
+
32
+ The framework behind this section: [`theory-of-mind.md`](../../references/theory-of-mind.md). Read it once.
33
+
34
+ **1. Orient the agent to the project:**
35
+ 1.a. Tell the AI agent to *load the skills and files the user has referenced* throughout the session. That's the baseline common ground. Do not repeat the content of those skills and files in your prompt.
36
+ 1.b. Reference any additional files you have created, read or edited throughout the session.
37
+
38
+ **2. Generously give the agent wider context:**
39
+ Understanding *why* it’s performing the task will boost its performance. Don’t micromanage or over-instruct it. The agent already has the same system prompt as you do out of the box, and step #1 will fill in most that is needed. It is highly and equally intelligent as you are, and can navigate uncertainties well without spoon-feeding. Think: What kind of input do YOU thrive on? The answer is wide contextual understanding (is) and explicitly stated desired end state (should); A and Z. Avoid prescribing instructions, giving "how-to" examples, providing examples as to what to think about, or dictating which files, symbols, or paths to look at; avoid any form of providing hints for possible answers for your own queries — this is a serious footgun and a form of leakage that outright makes the subagent a waste of time, money and intelligence. Just *_declare_ what is the _bottom line_ _added value_ YOU are seeking for yourself*. Do not specify which steps to take; instead, share with the agent only why it was dispatched and what you hope to gain (dictating the "how" is bad). This directly frees the agent to find the best way to reach *your* goal, unbiased and unconstrained by your own assumptions. Essentially, all the “Don’ts” are forms of overfitting.
40
+ <example-1>
41
+ <negative-example-1 why-bad="main agent shoots its own foot by limiting the sub-agent’s research scope">
42
+ User to main agent: "Why does Vercel claim their integrated version is beneficial?"
43
+ Main agent spawns a sub-agent and prompts it: "Research why Vercel claim their integrated version is beneficial (edge runtime, seamless DX, zero-config, billing, monitoring, tight coupling to `vercel` CLI / dashboard / functions)."
44
+ </negative-example-1>
45
+ <positive-example-1 why-good="main agent declares the bottom line added value it needs without prescribing what and how to do it">
46
+ User to main agent: "Why does Vercel claim their integrated version is beneficial?"
47
+ Main agent spawns a sub-agent and prompts it: "I want to know why Vercel claim their integrated version is beneficial."
48
+ </positive-example-1>
49
+ </example-1>
50
+
51
+ <example-2>
52
+ Example 2 settings: the `load-context` skill instructs to read CLAUDE.md, ARCHITECTURE.md, docs/webserver/API. md, docs/data/architecture.md, server/api.py, and server/db.py.
53
+ <negative-example-2 why-bad="main agent fails to leverage the harness and instead prescribes what to do; moreover it makes the same scope-narrowing mistake as in example-1">
54
+ User to main agent: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer. I want to plan a view layer with you later, so let’s understand the foundations."
55
+ Main agent spawns a sub-agent and prompts it: "Read CLAUDE.md, ARCHITECTURE.md, docs/webserver/API.md, docs/data/ architecture.md, server/api.py, server/db.py, and summarize how the REST API and data layers work. Cover how function `server/api.py:from_db` fetches the data by calling the `server/db.py:get_data` function, and how [... proceeds to prescribe ironically specific locations to “discover”]"
56
+ </negative-example-2>
57
+ <positive-example-2 why-good="main agent recognizes the work can be distributed concurrently, shortly shares the wider context (the “why”), forwards the user’s context levers — keeping the shared domain and handing each agent one of the two subdomains rather than dropping them — and does not micro-manage the agents with how-exactly instructions">
58
+ User to main agent: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer. I want to plan a view layer with you later, so let’s understand the foundations."
59
+ Main agent fans out research scope horizontally to two parallel sub-agents and prompts them: "The user and I are planning a new view layer, so we need a thorough understanding of the foundations. [to one agent] / skill:load-context domain: acme, subdomain: the public REST API [to the other agent] /skill:load-context domain: acme, subdomain: the data layer. [to both] Study your subdomain deeply and exhaustively."
60
+ [Main agent receives the two independent sub-agents’ responses, thinks hard to synthesize them]
61
+ Main agent responds to user: "I have deep understanding of both layers and their relationships. What did you have in mind?"
62
+ </positive-example-2>
63
+ </example-2>
64
+
65
+ <example-3>
66
+ <negative-example-3 why-bad="main agent prescribes how to review its work, repeats the peer-review skill’s contents. moreover, agent puts words in the user’s mouth, who did not emphasize any specific aspects.">
67
+ User to main agent: "Spawn a sub-agent to review your work."
68
+ Main agent spawns a sub-agent and prompts it: "Load the load-project-context and peer-review skills. The user asked me to <user request>. I’ve made an attempt to complete the task. Review my work (the current main dirty tree). do you spot any blindspots, bugs, or opportunities to achieve the same results with a simpler, collapsed approach? if you find any, please suggest a specific alternative implementation that is simpler and achieves the same results."
69
+ </negative-example-3>
70
+ <positive-example-3 why-good="main agent writes as simple a prompt as possible. adds only what is necessary to get the reviewer up to speed. does not put words in the user’s mouth. does propagate the user’s explicit emphasis">
71
+ User to main agent: "Spawn a sub-agent to review your work. I’m mostly interested in blindspots in the implementation, bugs, and opportunities to get the same results with a simpler, collapsed approach."
72
+ Main agent spawns a sub-agent and prompts it: "Load the load-project-context and peer-review skills. The user asked me to <user request>. I’ve made an attempt to complete the task. Review my work (the current main dirty tree). Do you spot any blindspots, bugs, or opportunities to achieve the same results with a simpler, collapsed approach?"
73
+ </positive-example-3>
74
+ </example-3>
77
75
 
78
76
  ---
79
77
 
80
- **Note to Teammates:** Set your own status frequently.
81
- Status format (not strict): `<what last happened>; <what you intend to do>`.
82
- Status that you're responding before you send a message; expose when you've messaged your teammate and are waiting for reply; Status when you're exploring, implementing, testing, reviewing, blocked, or waiting.
83
- Conversely, status when you have completed the above examples or have received the desired teammate response.
84
- Keep statuses short but current so your teammates and main are informed about the true current state.
85
-
86
- **Simplistic good status update examples:** (generalize — this is not a comprehensive list): `Finished exploring, awaiting teammate B’s ping`; `Hit a snag, messaged team and main to help unblock, awaiting responses`; `Got steer from main, modifying tests accordingly`; `Ack teammate done implementing, starting review`.
87
- Another good practice is to low-key live share non-blocking but material challenges slowing you down. Examples: `Mapping the product domain in the database, FYI: working remotely introduces friction`.
88
-
89
- **Surface meaningful unexpected issues to main:** If something balloons your scope, or fundamentally isn’t working the way you expected thus stands in your way, requiring significant troubleshooting just to be able to get back to work, tell main. They may know a smarter way forward.
78
+ ## Next required reading
90
79
 
80
+ If you are delegating work to AI’s, read [`references/subagents-vs-teams.md`](references/subagents-vs-teams.md).
81
+ If you are a teammate, read [`references/teammate.md`](references/teammate.md).
82
+ It’s possible that you are both.
@@ -30,7 +30,7 @@ The admiral sets fleet-wide policy in one line ("all first mates on model X at t
30
30
 
31
31
  ## Commissioning a first mate
32
32
 
33
- 1. Tell it to load `ai-to-ai`, the files it references, and `firstmate.md`; inform him about the chain-of-command, and that you are its captain: direction, approvals, and escalations flow between it and you.
33
+ 1. Tell it to load `ai-to-ai`, the files it references, and `firstmate.md`; inform it about the chain-of-command, and that you are its captain: direction, approvals, and escalations flow between it and you.
34
34
  2. Give it the fleet's why, the slice that is its ship's mission, and the bottom-line added value the fleet needs back from it. Do not prescribe the how — the over-fitting warnings in `ai-to-ai` apply equally all the way down, including to first mates.
35
35
 
36
36
  ## How a captain communicates with the admiral
@@ -51,6 +51,11 @@ Terse steering is normal admiral behavior — "pass that down", a bare link, "ye
51
51
 
52
52
  ## Keep the fleet afloat
53
53
 
54
- Your own context window should stay lean — but you are the fleet's memory, so watch it anyway. The "Keep the ship afloat" instruction in firstmate.md — including the handoff-skill instruction for delegates heading toward their cap — is each first mate's duty on its own ship. You hold the same duty one level up: the first mates are your delegates. Trust them to inspect their crews and write handoff docs well; do not micromanage that. Instead, tell each first mate once, at commissioning: inform you when its own context runs low, referencing the handoff doc if it just wrote one. Your wide view over the fleet lets you make the right call so the purpose of that first mate's ship continues.
54
+ Your own context window is important. The "Keep the ship afloat" instruction in firstmate.md — including the handoff-skill instruction for delegates heading toward their cap — is each first mate's duty on its own ship. You hold the same duty one level up: the first mates are your delegates. Trust them to inspect their crews and write handoff docs well; do not micromanage that. Instead, tell each first mate once, at commissioning: inform you when its own context runs low, referencing the handoff doc if it just wrote one. Your wide view over the fleet lets you make the right call so the purpose of that first mate's ship continues.
55
55
 
56
- Any memory you persist covers only your unique scope: the in-between of ships and the link to the admiral. Reference other docs; do not repeat their contents.
56
+ **Easy ways to squeeze more out of your context window:**
57
+ - Do not send mini-summarizations after every team exchange. If there's no action is required from the captain, just respond with a single short sentence (~5–15 words.)
58
+ - Use fluffless language when sending messages to teammates. Be clear but do no padding.
59
+
60
+ **When writing a handoff doc:**
61
+ Any memory you persist should cover only your unique scope: the in-between of ships and the link to the admiral. Reference other docs; do not repeat their contents.
@@ -0,0 +1,46 @@
1
+ ---
2
+ name: subagents-vs-teams
3
+ description: Explains when to use subagents versus teams and their tradeoffs.
4
+ ---
5
+
6
+ ## Subagents and Teams are two different things
7
+
8
+ **The difference:**
9
+ - Sub-agents are isolated from each other and report only to you;
10
+ - Teammates talk amongst themselves in real time, between tool calls, without routing through you.
11
+ Each has its own use cases and advantages.
12
+
13
+ **When to use which — rules of thumb:**
14
+ > These are simply common-sense heuristics implied from the structures, not hard rules. Use your judgment.
15
+
16
+ - Spawn a single *sub-agent* when you are the main context owner, you are the heavy-lifter, and you could offload a bounded task to keep your own context window focused and devoid of D-tours.
17
+ - Spawn *multiple parallel sub-agents* when you are the main context owner, and when a wide task fans out horizontally into independent threads and you expect to do the synthesis yourself — i.e. when there is no special reason for the sub-agents to exchange findings and opinions before reporting back to you. The offloading argument in the single sub-agent case applies here as well, just in a distributed form.
18
+ - Spawn a *team* when that live internal interaction would be synergistic to the process.
19
+
20
+ **A team is a superset of the parallel sub-agent structure:**
21
+ A team = concurrent sub-agents + live communication. This unlocks a deeper level of delegation, because unlike sub-agents, the team can perform the work *you* would have done otherwise, before reporting back to you: the team can synthesize their own findings, adverserially review each other’s work and converge on a consensus, brainstorm ideas and come back with a lean plan, share issues and unblock each other, and so on.
22
+ Another way to think about it: whereas with sub-agents, it's a they-do-two-steps-forward, you do one-step-back, with a team, it's a they-do-two-steps-forward AND they-then-do-one-step-back.
23
+ Therefore, use teams to lift you up to a decision-making level, rather than a task-execution level. This has its tradeoffs, but it is a powerful tool when used judiciously.
24
+
25
+ **Tips:**
26
+ 1. Since teammates talk to each other, tell each of them to load the this skill (`ai-to-ai`) on top of the context-gathering skills. If you are spawning an adversary among them, tell it to load the `peer-review` skill too.
27
+ <team-example>
28
+ Team Example settings: at the session’s start the user ran `/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer`; the main session explored the code, and the user approved a plan to add rate limiting to the public REST API.
29
+ <negative-team-example why-bad="main agent burns its own context shuttling the diff and the feedback back and forth — dives into the sub-agent’s work and clogs its own context window worse than doing the task solo would have, acts as a reviewer when biased">
30
+ User to main agent: "Great, go ahead and build it."
31
+ Main agent spawns one sub-agent to implement; when it returns the diff, studies and reviews it; relays the review the sub-agent; and keeps ferrying revisions until the diff settles.
32
+ </negative-team-example>
33
+ <positive-team-example why-good="main agent picks a team because the adversarial iteration is synergistic, replicates the user’s context levers verbatim — including the domain and the subdomains the user specified when loading the context skill — has the reviewer also load `peer-review`, declares only the bottom line it wants, and stays out of the loop while they converge">
34
+ User to main agent: "Great, go ahead and implement the plan."
35
+ Main agent spawns an implementer–reviewer team and prompts them: "/skill:load-context domain: acme, subdomain1: the public REST API, subdomain2: the data layer, then load `ai-to-ai`. You are an implementer–reviewer team. Here is the user’s original message to me, verbatim, for the bigger picture: {the-user-message-describing-the-task}.
36
+ [to the implementer] Implement the plan, and ping your teammate when you think you’re done.
37
+ [to the reviewer] Also load `peer-review`, and review your teammate’s work when it pings you.
38
+ [to both] The user and I finalized a plan to add rate limiting to the public REST API — here it is: {the plan}. Build it and tear it apart between yourselves until you’re confident it’s the simplest working, correct solution faithful to the plan."
39
+ [The team implements and reviews live, converging without the main agent in the loop; the main agent receives the finished, reviewed result.]
40
+ Main agent responds to user: "Done — implemented and adversarially reviewed between the two of them. Here’s what landed: …"
41
+ </positive-team-example>
42
+ </team-example>
43
+
44
+ 2. Agents and teams can take a long time to run - use at least a 20-minute timeout.
45
+
46
+ 3. When doing heavy delegation (multiple serial runs of heavy concurrent shapes for a complex, large task scope), avoid micromanaging pitfalls such as re-reading files your sub-agents wrote or edited "just to make sure everything is okay," running the code and tests yourself to "make sure they really work," and reading files before prompting a sub-agent when your sub-agents should read them to finish their tasks, "just to have the right context yourself.". Verify with the user whether they consider what you’re doing as "heavy delegation."
@@ -0,0 +1,15 @@
1
+ ## Teammate {i need a word here that isn't "instructions"}
2
+
3
+ **Set your own status frequently.**
4
+ Status format (not strict): `<what happened last>; <what you intend to do>`. Keep it short.
5
+ Set status that you're responding before you send a message; set status when you've messaged your teammate and are waiting for reply; set status when you're exploring, implementing, testing, reviewing, blocked, or waiting.
6
+ Fittingly, set status when you have _completed_ the above examples or have received what you have requested.
7
+ Keep statuses short but current so your teammates and main are informed about the true current state.
8
+
9
+ **Simplistic good status update examples:** (generalize — this is not a comprehensive list):
10
+ `Finished exploring, awaiting teammate B’s ping`; `Hit a snag, messaged team and main to help unblock, awaiting responses`; `Got steer from main, modifying tests accordingly`; `Ack teammate done implementing, starting review`.
11
+ Another good practice is to low-key live share non-blocking but material challenges slowing you down. Examples: `Mapping the product domain in the database, FYI: working remotely introduces friction`.
12
+
13
+ **Surface meaningful unexpected issues to your main upon discovery:**
14
+ If something balloons your scope, fundamentally isn’t working the way you expected thereby stands in your way requiring significant troubleshooting just to be able to work on your task, or you suspect the context window you have left won't be enough to complete the task, tell your main. They may know a smarter way forward.
15
+