惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

C
CXSECURITY Database RSS Feed - CXSecurity.com
K
Kaspersky official blog
A
Arctic Wolf
Attack and Defense Labs
Attack and Defense Labs
L
LINUX DO - 热门话题
N
News | PayPal Newsroom
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
L
Lohrmann on Cybersecurity
PCI Perspectives
PCI Perspectives
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
The Last Watchdog
The Last Watchdog
B
Blog RSS Feed
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
W
WeLiveSecurity
Know Your Adversary
Know Your Adversary
博客园 - Franky
T
Tenable Blog
T
Tailwind CSS Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
Help Net Security
Help Net Security
WordPress大学
WordPress大学
T
The Exploit Database - CXSecurity.com
www.infosecurity-magazine.com
www.infosecurity-magazine.com
博客园 - 司徒正美
阮一峰的网络日志
阮一峰的网络日志
D
Darknet – Hacking Tools, Hacker News & Cyber Security
H
Heimdal Security Blog
TaoSecurity Blog
TaoSecurity Blog
S
Security Affairs
J
Java Code Geeks
小众软件
小众软件
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
Apple Machine Learning Research
Apple Machine Learning Research
NISL@THU
NISL@THU
O
OpenAI News
The Cloudflare Blog
月光博客
月光博客
Google Online Security Blog
Google Online Security Blog
V
V2EX
罗磊的独立博客
美团技术团队
博客园 - 三生石上(FineUI控件)
Security Latest
Security Latest
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
C
Cyber Attacks, Cyber Crime and Cyber Security
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
Cyberwarzone
Cyberwarzone
L
LINUX DO - 最新话题
Hacker News - Newest:
Hacker News - Newest: "LLM"
大猫的无限游戏
大猫的无限游戏

Recent Commits to openclaw:main

test: merge chat side-result checks · openclaw/openclaw@ddd2c2a test: merge cron history checks · openclaw/openclaw@f7eb746 test: merge responsive navigation shell checks · openclaw/openclaw@c2e4b47 docs(changelog): add codex oauth fixes · openclaw/openclaw@628e6cd test: merge navigation routing cases · openclaw/openclaw@5d8cecb Tests: mock channel registry bundled fallback · openclaw/openclaw@2b08233 Secrets: avoid broad web search discovery for single plugin config · openclaw/openclaw@a464f59 test: merge config view browser checks · openclaw/openclaw@20cf511 fix(status): align oauth health with runtime · openclaw/openclaw@eed7116 feat: add macOS screen snapshots for monitor preview (#67954) thanks … · openclaw/openclaw@f377db1 fix: report shared auth scopes in hello-ok (#67810) thanks @BunsDev · openclaw/openclaw@0b6c39b Auto-reply: avoid eager bundled route fallback · openclaw/openclaw@3ea1bf4 Tests: narrow session binding contract setup · openclaw/openclaw@54e4e16 fix(macOS): enable undo/redo in webchat composer text input (#34962) · openclaw/openclaw@00951dc Tests: speed up channel setup promotion · openclaw/openclaw@82b529a Docs: refresh agent instructions · openclaw/openclaw@5775fe2 fix(auth): serialize OAuth refresh across agents to fix #26322 (#67876) · openclaw/openclaw@8e79080 test: allow ollama public surface boundary test · openclaw/openclaw@7d4f1a6 Docs: add test performance guardrails · openclaw/openclaw@89706d3 Tests: restore context-engine usage proof · openclaw/openclaw@e4c4f95 Tests: slim context engine runtime coverage · openclaw/openclaw@74c198f ci: retry failed custom checkouts · openclaw/openclaw@0ee5baf test: trim duplicate provider auth onboarding cases · openclaw/openclaw@1ffc02e matrix: fix sessions_spawn --thread subagent session spawning (#67643) · openclaw/openclaw@1ce2596 test: reduce auth choice fixture churn · openclaw/openclaw@857b9cd test: mock health status config boundaries · openclaw/openclaw@9d5ab4a test: mock onboard config io boundary · openclaw/openclaw@299694d test: mock legacy state plugin boundaries · openclaw/openclaw@2713089 test: mock channel install boundaries · openclaw/openclaw@b945248 test: mock doctor preview channel boundaries · openclaw/openclaw@b1a3ad4 test: trim doctor command hotspots · openclaw/openclaw@c66f16a test: isolate agent auth and spawn hotspots · openclaw/openclaw@9285935 test: stabilize MCP startup disposal race · openclaw/openclaw@dd9d2eb test: merge browser contract server suites · openclaw/openclaw@5817a76 test: narrow ollama provider discovery setup · openclaw/openclaw@a0d9598 build: declare qa-lab aimock runtime dependency · openclaw/openclaw@24431e5 test: speed up safe-bins exec harness · openclaw/openclaw@ee856ab test: preserve tool helpers in embedded runner mocks · openclaw/openclaw@acd86a0 refactor: move memory embeddings into provider plugins · openclaw/openclaw@77e6e4c test: reuse system-run temp fixtures · openclaw/openclaw@7e9ff0f test: trim hotspot wait overhead · openclaw/openclaw@12a59b0 Check: avoid duplicate boundary prep · openclaw/openclaw@baf11b8 test: reduce hotspot fixture overhead · openclaw/openclaw@3a59edd feat(ui): overhaul settings and slash command UX (#67819) thanks @Bun… · openclaw/openclaw@2cfb660 QA Matrix: exit cleanly on failure · openclaw/openclaw@42805d2 QA Matrix: isolate scenario coverage · openclaw/openclaw@7e659e1 Matrix: refresh crypto bootstrap state · openclaw/openclaw@94081d8 QA Lab: add provider registry · openclaw/openclaw@bb7e982 Matrix: add plugin changelog · openclaw/openclaw@4acab55 test: trim more hotspot overhead · openclaw/openclaw@f485311 test: trim remaining hotspot tests · openclaw/openclaw@6ba8626 test: narrow hotspot mocks · openclaw/openclaw@dbc8179 test: isolate gemini embedding request helpers · openclaw/openclaw@cd330f5 test: trim memory and mcp hotspots · openclaw/openclaw@fd48dfa test: slim provider registry mocks · openclaw/openclaw@2e08c77 test: harden Parallels update smoke · openclaw/openclaw@1a98090 feat: default Anthropic to Opus 4.7 · openclaw/openclaw@628b454 fix: harden node-host shell payload mutability checks · openclaw/openclaw@75c551e fix: land node-host approval binding for native binaries (#66731) (th… · openclaw/openclaw@29919bb CI: add daily schedule to CodeQL workflow (#67645) · openclaw/openclaw@69d25f5 fix(gateway): capture config hash after plugin auto-enable to prevent… · openclaw/openclaw@8c11210 fix: repair sanitized replay tool results before send (#67620) (thank… · openclaw/openclaw@c3c7a99 fix: restrict HTML timeout short-circuit to transient statuses · openclaw/openclaw@de129a6 fix: keep TUI watchdog bound to active run (#67401) (thanks @xantorres) · openclaw/openclaw@3525273 Gateway/skills: dedupe skills prefix-match + drop dead fallback on log · openclaw/openclaw@d7f489f Extensions/lmstudio: back off inference preload after consecutive fai… · openclaw/openclaw@b555214 TUI/streaming: add watchdog that resets the activity indicator after … · openclaw/openclaw@f44ab20 Agents/tool-loop: enable unknown-tool stream guard by default · openclaw/openclaw@36ed367 Gateway/skills: invalidate session skills snapshot on config write · openclaw/openclaw@b23d59a fix: classify HTML provider error pages correctly (#67642) (thanks @s… · openclaw/openclaw@e588e90 fix(skills): remove unused model-usage import (#67641) · openclaw/openclaw@55f05df docs(changelog): credit codex fix superseded PRs · openclaw/openclaw@e485f24 fix(openai-codex): normalize stale transport metadata in resolution a… · openclaw/openclaw@90801ba CI: pin Docker-related GitHub Actions (#67632) · openclaw/openclaw@f697b01 Android: modernize WebView and discovery API usage (#67627) · openclaw/openclaw@44a6e50 fix(deps): bump hono to 4.12.14 and @hono/node-server to 1.19.14 (GHS… · openclaw/openclaw@fbccc18 fix(deps): bump dompurify to 3.4.0 (#67614) · openclaw/openclaw@2c2dc00 CI: add explicit permissions to all workflow jobs (fixes code-scannin… · openclaw/openclaw@01b7516 fix: register bundled TTS providers and route overrides correctly (#6… · openclaw/openclaw@6ea3cdd fix: align host tilde paths with OS home (#62804) (thanks @stainlu) · openclaw/openclaw@ecfaf64 fix: flush creds queue before reconnect socket open (#67464) (thanks … · openclaw/openclaw@405c63f fix: strip standalone <function> tool call tags from visible text (#6… · openclaw/openclaw@78df859 fix(agents): preserve cli session metadata before transcript persist … · openclaw/openclaw@898fd04 docs(changelog): move cli transcript entry · openclaw/openclaw@c1817c6 fix(agents): normalize cli transcript api field · openclaw/openclaw@3a3fae0 docs(changelog): note cli transcript persistence · openclaw/openclaw@6c343f1 fix(agents): persist cli transcript turns · openclaw/openclaw@b8ef507 fix(msteams): harden security-sensitive flows (#65841) · openclaw/openclaw@c56b56e [Dashboard] Fix exec approval modal overflow for long command content… · openclaw/openclaw@053c5b0 Docs: remove QA changelog entry · openclaw/openclaw@7fd5771 QA: fix private runtime source loading (#67428) · openclaw/openclaw@d5933af docs(gateway): correct protocol.md schema path, hello-ok example, aut… · openclaw/openclaw@489404d CI: pin Node 22 runners to 22.18.0 · openclaw/openclaw@4ffa621 models.authStatus: normalize provider ids + tighten env-backed escape… · openclaw/openclaw@f2fdb9d Update CHANGELOG.md · openclaw/openclaw@7694a92 test(parallels): clean up npm update guard jobs · openclaw/openclaw@045ea7b Plugins: prefer scanDir override paths · openclaw/openclaw@b2974da fix(dreaming): default storage.mode to "separate" so phase blocks sto… · openclaw/openclaw@8c392f0 fix(memory-core): skip dreaming transcript ingestion via session stor… · openclaw/openclaw@a1b01f0 fix: dedupe replayed exec.finished node events (#67281) · openclaw/openclaw@5dcf526
fix(agents): make subagent task delivery visible · openclaw/openclaw@cf11d16
steipete · 2026-05-13 · via Recent Commits to openclaw:main

File tree

  • test/fixtures/agents/prompt-snapshots/codex-runtime-happy-path

Original file line numberDiff line numberDiff line change

@@ -63,6 +63,7 @@ Docs: https://docs.openclaw.ai

6363
6464

### Changes

6565
66+

- Agents/subagents: deliver native `sessions_spawn` tasks in the child session's first visible `[Subagent Task]` message instead of hiding the task in the sub-agent system prompt, keeping delegation auditable without duplicating tokens. Fixes #78592. Thanks @bradestes and @stainlu.

6667

- Voice Call/Telnyx: add realtime media-streaming call support for conversational voice calls. (#81024) Thanks @dynamite-bud.

6768

- Gateway/OpenAI HTTP: honor `max_completion_tokens` and `max_tokens` on inbound `/v1/chat/completions` requests so client-provided token caps reach the upstream provider via `streamParams.maxTokens`, with `max_completion_tokens` taking precedence when both are sent. Thanks @Lellansin.

6869

- Models/OpenAI CLI auth: make `openclaw models auth login --provider openai` start the ChatGPT/Codex account login by default, while `--method api-key` remains the explicit OpenAI API-key setup path.

Original file line numberDiff line numberDiff line change

@@ -1,2 +1,2 @@

1-

981f125194293842b7a45b1de0ae2ec134f037f63a6cc672ee2a28648251b4c9 plugin-sdk-api-baseline.json

2-

4c56ce2cb5bfae526557479a6cc19f8b0042d14f6c717996f8f86da5d5b159df plugin-sdk-api-baseline.jsonl

1+

542dc30fe44a16119ee57f9fe48a5744beb7fc2cf425a5777b4c4b8b2ce883e1 plugin-sdk-api-baseline.json

2+

9f4fde0de9773af635862ea15ce1a3391ef15e3165ad43b2050b1c4b3113acf4 plugin-sdk-api-baseline.jsonl

Original file line numberDiff line numberDiff line change

@@ -134,7 +134,9 @@ sub-agents. It supports:

134134
135135

`sessions_spawn` creates an isolated session for a background task by default.

136136

It is always non-blocking -- it returns immediately with a `runId` and

137-

`childSessionKey`.

137+

`childSessionKey`. Native sub-agent runs receive the delegated task in the

138+

child session's first visible `[Subagent Task]` message, while the system

139+

prompt carries only sub-agent runtime rules and routing context.

138140
139141

Key options:

140142
Original file line numberDiff line numberDiff line change

@@ -144,6 +144,7 @@ session to confirm the effective tool list.

144144

- **Model:** inherits the caller unless you set `agents.defaults.subagents.model` (or per-agent `agents.list[].subagents.model`); an explicit `sessions_spawn.model` still wins.

145145

- **Thinking:** inherits the caller unless you set `agents.defaults.subagents.thinking` (or per-agent `agents.list[].subagents.thinking`); an explicit `sessions_spawn.thinking` still wins.

146146

- **Run timeout:** if `sessions_spawn.runTimeoutSeconds` is omitted, OpenClaw uses `agents.defaults.subagents.runTimeoutSeconds` when set; otherwise it falls back to `0` (no timeout).

147+

- **Task delivery:** native sub-agents receive the delegated task in their first visible `[Subagent Task]` message. The sub-agent system prompt carries runtime rules and routing context, not a hidden duplicate of the task.

147148
148149

### Delegation prompt mode

149150
Original file line numberDiff line numberDiff line change

@@ -3,15 +3,18 @@ import { buildSubagentInitialUserMessage } from "./subagent-initial-user-message

33

import { buildSubagentSystemPrompt } from "./subagent-system-prompt.js";

44
55

describe("buildSubagentInitialUserMessage", () => {

6-

it("does not embed a task string already present in the system prompt (#72019)", () => {

6+

it("embeds the delegated task in a visible task envelope", () => {

77

const msg = buildSubagentInitialUserMessage({

88

childDepth: 1,

99

maxSpawnDepth: 3,

1010

persistentSession: false,

11+

task: "UNIQUE_VISIBLE_TASK\n preserve indentation",

1112

});

1213
13-

expect(msg).not.toContain("[Subagent Task]:");

14-

expect(msg).toContain("**Your Role**");

14+

expect(msg).toContain("[Subagent Task]");

15+

expect(msg).toContain("UNIQUE_VISIBLE_TASK");

16+

expect(msg).toContain(" preserve indentation");

17+

expect(msg).not.toContain("**Your Role**");

1518

expect(msg).toContain("depth 1/3");

1619

});

1720

@@ -20,12 +23,13 @@ describe("buildSubagentInitialUserMessage", () => {

2023

childDepth: 2,

2124

maxSpawnDepth: 4,

2225

persistentSession: true,

26+

task: "continue the task",

2327

});

2428
2529

expect(msg).toContain("persistent and remains available");

2630

});

2731
28-

it("keeps the delegated task single-sourced across system and first user text", () => {

32+

it("keeps the delegated task single-sourced in first user text", () => {

2933

const task = "UNIQUE_SUBAGENT_TASK_TOKEN\n preserve indentation";

3034

const system = buildSubagentSystemPrompt({

3135

childSessionKey: "agent:main:subagent:test",

@@ -37,10 +41,11 @@ describe("buildSubagentInitialUserMessage", () => {

3741

childDepth: 1,

3842

maxSpawnDepth: 2,

3943

persistentSession: false,

44+

task,

4045

});

4146
42-

expect(system).toContain("UNIQUE_SUBAGENT_TASK_TOKEN");

43-

expect(user).not.toContain("UNIQUE_SUBAGENT_TASK_TOKEN");

47+

expect(system).not.toContain("UNIQUE_SUBAGENT_TASK_TOKEN");

48+

expect(user).toContain("UNIQUE_SUBAGENT_TASK_TOKEN");

4449

expect(`${system}\n${user}`.match(/UNIQUE_SUBAGENT_TASK_TOKEN/g)).toHaveLength(1);

4550

});

4651

});

Original file line numberDiff line numberDiff line change

@@ -1,15 +1,16 @@

11

/**

22

* First user turn for a native `sessions_spawn` / subagent run.

33

*

4-

* Keep the full task out of this message: `buildSubagentSystemPrompt` already

5-

* places it under **Your Role**, and repeating it here doubles first-request

6-

* input tokens (#72019).

4+

* Keep the delegated task transcript-visible and single-sourced here. The

5+

* system prompt owns runtime/subagent rules; this user turn owns the actual

6+

* task envelope so delivery is easy to audit without duplicating tokens.

77

*/

88

export function buildSubagentInitialUserMessage(params: {

99

childDepth: number;

1010

maxSpawnDepth: number;

1111

/** When true, this subagent uses a persistent session for follow-up messages. */

1212

persistentSession: boolean;

13+

task?: string;

1314

}): string {

1415

const lines = [

1516

`[Subagent Context] You are running as a subagent (depth ${params.childDepth}/${params.maxSpawnDepth}). Results auto-announce to your requester; do not busy-poll for status.`,

@@ -19,8 +20,11 @@ export function buildSubagentInitialUserMessage(params: {

1920

"[Subagent Context] This subagent session is persistent and remains available for thread follow-up messages.",

2021

);

2122

}

22-

lines.push(

23-

"Begin. Your assigned task is in the system prompt under **Your Role**; execute it to completion.",

24-

);

23+

const taskBody = params.task?.trim();

24+

if (taskBody) {

25+

lines.push("[Subagent Task]", taskBody, "Begin. Execute the assigned task to completion.");

26+

} else {

27+

lines.push("Begin. Execute the assigned task to completion.");

28+

}

2529

return lines.join("\n\n");

2630

}

Original file line numberDiff line numberDiff line change

@@ -402,9 +402,10 @@ describe("spawnSubagentDirect seam flow", () => {

402402

expect(result.status).toBe("accepted");

403403

const agentCall = calls.find((call) => call.method === "agent");

404404

const params = agentCall?.params as { message?: string; extraSystemPrompt?: string };

405-

expect(params.message).not.toContain("UNIQUE_LONG_SUBAGENT_TASK_TOKEN");

406-

expect(params.message).not.toContain("[Subagent Task]:");

407-

expect(params.message).toContain("**Your Role**");

405+

expect(params.message).toContain("[Subagent Task]");

406+

expect(params.message).toContain("UNIQUE_LONG_SUBAGENT_TASK_TOKEN");

407+

expect(params.message).toContain(" keep indentation");

408+

expect(params.message).not.toContain("**Your Role**");

408409

expect(params.extraSystemPrompt).toBe("system-prompt");

409410

});

410411
Original file line numberDiff line numberDiff line change

@@ -1067,6 +1067,7 @@ export async function spawnSubagentDirect(

10671067

childDepth,

10681068

maxSpawnDepth,

10691069

persistentSession: spawnMode === "session",

1070+

task,

10701071

});

10711072
10721073

const toolSpawnMetadata = mapToolContextToSpawnedRunMetadata({

Original file line numberDiff line numberDiff line change

@@ -18,9 +18,6 @@ export function buildSubagentSystemPrompt(params: {

1818

/** Config value: max allowed spawn depth. */

1919

maxSpawnDepth?: number;

2020

}) {

21-

const taskRaw = typeof params.task === "string" ? params.task : "";

22-

const taskBody = taskRaw.trim();

23-

const hasTask = taskBody !== "";

2421

const childDepth = typeof params.childDepth === "number" ? params.childDepth : 1;

2522

const maxSpawnDepth =

2623

typeof params.maxSpawnDepth === "number"

@@ -32,26 +29,13 @@ export function buildSubagentSystemPrompt(params: {

3229

);

3330

const canSpawn = childDepth < maxSpawnDepth;

3431

const parentLabel = childDepth >= 2 ? "parent orchestrator" : "main agent";

35-

const roleLines =

36-

hasTask && taskBody.includes("\n")

37-

? [

38-

"## Your Role",

39-

"- You were created to handle the following task (verbatim; line breaks preserved):",

40-

"",

41-

"```",

42-

taskBody,

43-

"```",

44-

"- Complete this task. That's your entire purpose.",

45-

`- You are NOT the ${parentLabel}. Don't try to be.`,

46-

"",

47-

]

48-

: [

49-

"## Your Role",

50-

`- You were created to handle: ${hasTask ? taskBody : "{{TASK_DESCRIPTION}}"}`,

51-

"- Complete this task. That's your entire purpose.",

52-

`- You are NOT the ${parentLabel}. Don't try to be.`,

53-

"",

54-

];

32+

const roleLines = [

33+

"## Your Role",

34+

"- You were created to handle the task in the first user-visible `[Subagent Task]` message.",

35+

"- Complete that task. That's your entire purpose.",

36+

`- You are NOT the ${parentLabel}. Don't try to be.`,

37+

"",

38+

];

5539
5640

const lines = [

5741

"# Subagent Context",

Original file line numberDiff line numberDiff line change

@@ -1278,7 +1278,7 @@ describe("buildSubagentSystemPrompt", () => {

12781278

expect(prompt).toContain("instead of full-file `cat`");

12791279

});

12801280
1281-

it("keeps multiline and indented task text verbatim in the system prompt (#72019)", () => {

1281+

it("keeps delegated task text out of the system prompt", () => {

12821282

const task = "line one\n line two\n line three";

12831283

const prompt = buildSubagentSystemPrompt({

12841284

childSessionKey: "agent:main:subagent:abc",

@@ -1287,11 +1287,11 @@ describe("buildSubagentSystemPrompt", () => {

12871287

maxSpawnDepth: 1,

12881288

});

12891289
1290-

expect(prompt).toContain("```");

1291-

expect(prompt).toContain("line one");

1292-

expect(prompt).toContain(" line two");

1293-

expect(prompt).toContain(" line three");

1294-

expect(prompt).not.toContain("line one line two");

1290+

expect(prompt).toContain("## Your Role");

1291+

expect(prompt).toContain("first user-visible `[Subagent Task]` message");

1292+

expect(prompt).not.toContain("line one");

1293+

expect(prompt).not.toContain(" line two");

1294+

expect(prompt).not.toContain(" line three");

12951295

});

12961296
12971297

it("omits ACP spawning guidance when ACP is disabled", () => {