@11agents/cli 0.1.25 → 0.1.27

package/mobile-runtime/skills/android-collect-tiktok-metrics/SKILL.md

@@ -10,7 +10,7 @@ description: Collect TikTok account and video metrics from the bundled 11agents
 - Use `11agents mobile ...`; do not call raw ADB/Appium directly.
 - Collect only metrics visible to an authorized account/device.
 - Do not automate engagement, browsing, liking, following, sharing, or view inflation.
-- If TikTok Studio, login, verification, or app state requires a person, return `needs_human` with screenshot evidence.
+- If TikTok Studio, login, verification, or app state cannot be handled automatically, return `failed` with screenshot evidence.
 - Always pass `--json` when supported and `--task-id <task_id>` when running for an agent task.
 
 ## Required Inputs

package/mobile-runtime/skills/android-group-control-cli/SKILL.md

@@ -13,7 +13,7 @@ description: Operate the bundled 11agents mobile Android runtime for device-pool
 - Use only allowlisted high-level commands exposed by `11agents mobile`.
 - Do not run arbitrary `adb shell`, arbitrary Appium scripts, credential scraping, CAPTCHA bypass, login bypass, or platform verification bypass.
 - Do not automate likes, comments, follows, shares, views, random browsing, registrations, or fake engagement.
-- If a page needs human action, report `needs_human`, include the screenshot path, and stop.
+- If the CLI cannot continue automatically, report `failed`, include the screenshot path, and stop.
 - Prefer `--json` for machine-readable execution results.
 - Use `--dry-run` when the user asks for preview/test; use live publish only with explicit approval.
 

package/mobile-runtime/skills/android-group-control-cli/references/command-reference.md

@@ -118,5 +118,5 @@ Use `--no-ensure-appium` or `--no-appium-smoke` only for controlled debugging ag
 
 - `published`: report success, duration, record id, permalink/post id when available, screenshot, and run log path.
 - `dry_run`: report the post-form screenshot and say no publish was clicked.
-- `needs_human`: report screenshot and the reason; do not continue.
+- `failed`: report screenshot and the reason; do not continue.
 - `failed`: report screenshot and error; do not retry unless the user asks or the failure is a transient device/offline issue.

package/mobile-runtime/skills/android-publish-facebook/SKILL.md

@@ -34,7 +34,7 @@ Stop on login, restriction, verification, unexpected composer, or publish confir
 
 ## Output
 
-Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error or human action, and the data-collection handoff:
+Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform facebook --record-id <record_id> --json --task-id TASK_123

package/mobile-runtime/skills/android-publish-instagram/SKILL.md

@@ -34,11 +34,11 @@ Feed image or video:
 11agents mobile publish-instagram --device D03 --post-type post-video --video "/path/video.mp4" --caption "caption" --json --task-id TASK_123
 ```
 
-Do not auto-confirm unexpected login, verification, restriction, or account prompts. Stop on `needs_human` and return screenshot evidence.
+Do not auto-confirm unexpected login, verification, restriction, or account prompts. Stop on `failed` and return screenshot evidence.
 
 ## Output
 
-Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error or human action, and the data-collection handoff:
+Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform instagram --record-id <record_id> --json --task-id TASK_123

package/mobile-runtime/skills/android-publish-reddit/SKILL.md

@@ -34,7 +34,7 @@ Use `--no-fetch-permalink` when the public username is unknown and app-level pub
 
 ## Output
 
-Return publish status, record id, account/subreddit/device, screenshot path, permalink or platform post id when available, duration, error or human action, and the data-collection handoff:
+Return publish status, record id, account/subreddit/device, screenshot path, permalink or platform post id when available, duration, error, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform reddit --record-id <record_id> --json --task-id TASK_123

package/mobile-runtime/skills/android-publish-tiktok/SKILL.md

@@ -36,7 +36,7 @@ Use `--range`, `--parallel`, and `--max-concurrency N` only when the approved pl
 
 ## Output
 
-Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error or `needs_human` action, and the data-collection handoff:
+Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error action, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform tiktok --scope video --record-id <record_id> --json --task-id TASK_123

package/mobile-runtime/skills/android-publish-x/SKILL.md

@@ -33,7 +33,7 @@ Do not automate replies, follows, likes, reposts, random browsing, or DM actions
 
 ## Output
 
-Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error or human action, and the data-collection handoff:
+Return publish status, record id, account/device, screenshot path, permalink or platform post id when available, duration, error, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform x --record-id <record_id> --json --task-id TASK_123

package/mobile-runtime/skills/android-publish-xiaohongshu/SKILL.md

@@ -23,8 +23,8 @@ If approval, account lifecycle, media/package path, text fields, or device readi
 Image/video note:
 
 ```bash
-11agents mobile publish-xiaohongshu --device D03 --post-type image --image "/path/image.jpg" --title "title" --caption "body" --json --task-id TASK_123
-11agents mobile publish-xiaohongshu --device D03 --post-type video --video "/path/video.mp4" --title "title" --caption "body" --json --task-id TASK_123
+11agents mobile publish-xiaohongshu --device D03 --post-type image --image "/path/image.jpg" --title "title" --caption "body" --copy-link-after-publish --json --task-id TASK_123
+11agents mobile publish-xiaohongshu --device D03 --post-type video --video "/path/video.mp4" --title "title" --caption "body" --copy-link-after-publish --json --task-id TASK_123
 ```
 
 Structured package:
@@ -33,17 +33,19 @@ Structured package:
 11agents mobile publish-xiaohongshu --device D03 --publish-package "/path/publish_package.json" --json --task-id TASK_123
 ```
 
-Link copy after publish, when requested:
+Successful live publish is only confirmed after the CLI taps the final publish button, waits without tapping while Xiaohongshu shows upload/progress state, then opens the "我" page, opens the note whose title matches the expected title, and verifies the detail page against the expected title/body. If that evidence does not match, treat the result as `failed`; do not claim publish success from the publish button tap alone.
+
+Link copy after publish or during the recovery skill:
 
 ```bash
-11agents mobile copy-xiaohongshu-link --device D03 --record-id <record_id> --json --task-id TASK_123
+11agents mobile copy-xiaohongshu-link --record-id <record_id> --json --task-id TASK_123
 ```
 
-Use `--confirm-no-disclosure-needed` only when the approved package/policy explicitly allows it. Stop on login, restriction, verification, disclosure ambiguity, or unexpected UI state.
+Use `--confirm-no-disclosure-needed` only when the approved package/policy explicitly allows it. Return `failed` with screenshot evidence on login, restriction, verification, disclosure ambiguity, or unexpected UI state.
 
 ## Output
 
-Return publish status, record id, account/device, screenshot path, note link or platform post id when available, duration, error or human action, and the data-collection handoff:
+Return publish status, record id, account/device, screenshot path, note link or platform post id when available, duration, error, and the data-collection handoff:
 
 ```bash
 11agents mobile data collect --platform xiaohongshu --record-id <record_id> --copy-link --json --task-id TASK_123

package/mobile-runtime/skills/mobile-publish-device-health/SKILL.md

@@ -24,7 +24,7 @@ If mobile setup, device data, or target device information is missing, ask the u
 
 3. Select target devices only from current health output where the managed device is active and online.
 4. If account inventory was requested and `~/.11agents/mobile/data/accounts.csv` exists, check account binding through documented local commands or records. If devices are already logged in and no account registry was requested, verify app/login readiness from device state instead of requiring account metadata.
-5. Capture screenshots through high-level CLI when a device/account needs human attention:
+5. Capture screenshots through high-level CLI when a device/account needs inspection:
 
 ```bash
 11agents mobile screenshot Dxx --task-id TASK_ID
@@ -39,8 +39,8 @@ If mobile setup, device data, or target device information is missing, ask the u
 ```
 
 7. `tap`, `keyevent`, and `input-text` are allowed only through `11agents mobile` for explicit diagnostics, documented recovery steps, or human handoff assistance. Do not run raw `adb shell`.
-8. Do not return `needs_human` only for camera, microphone, gallery/media, or notification permission prompts when the publish CLI can auto-grant or tap allow for publish-required permissions.
-9. Return `needs_human` for login prompts, captcha, account restrictions, verification challenges, or unexpected app states that are not permission prompts.
+8. Do not fail only for camera, microphone, gallery/media, or notification permission prompts when the publish CLI can auto-grant or tap allow for publish-required permissions.
+9. Return `failed` for login prompts, captcha, account restrictions, verification challenges, or unexpected app states that are not permission prompts.
 10. Do not retry blindly after a restricted or unknown state; preserve screenshot evidence and stop.
 
 ## Output

package/mobile-runtime/skills/mobile-publish-execution/SKILL.md

@@ -47,11 +47,11 @@ If mobile setup, device online status, media path, app login readiness, or appro
 7. Validate the selected platform command supports the requested post type and text fields.
 8. If no explicit `--account-id` is provided, rely on the command's `--account-prefix` fallback account id for publish records when supported.
 9. Use `--dry-run` only for preview/test requests; do not add it to explicit live-publish requests.
-10. Do not stop solely for Android camera, microphone, gallery/media, notification, Reels public-sharing, or Xiaohongshu disclosure prompts when the CLI policy handles them for explicit live publish. Let the CLI return `needs_human` when it cannot safely continue.
+10. Do not stop solely for Android camera, microphone, gallery/media, notification, Reels public-sharing, or Xiaohongshu disclosure prompts when the CLI policy handles them for explicit live publish. Let the CLI return `failed` with screenshot evidence when it cannot safely continue automatically.
 11. Use `11agents mobile publish-tiktok`, `publish-instagram`, `publish-facebook`, `publish-reddit`, `publish-x`, and `publish-xiaohongshu`.
 12. Prefer `--json` and parse status, duration, record id, screenshot path, permalink, platform post id, and error.
 13. Pass `--task-id <task_id>` so logs land at `~/.11agents/mobile/runs/<task_id>/log`.
-14. Stop on `needs_human` or `failed`; surface screenshot and error instead of repeated retries.
+14. Stop on `failed`; surface screenshot and error instead of repeated retries.
 
 ## Output
 Return readiness status, dry-run result or execution preparation summary, device/app health notes, optional account-registry notes, command family selected, exact safe `11agents mobile` command shape, log path, and missing env placeholders.

package/mobile-runtime/skills/mobile-publish-records/SKILL.md

@@ -21,7 +21,7 @@ If record path, record id, publish result, or update intent is missing, ask the
 ```
 
 3. Update records only through high-level local commands such as `11agents mobile update-publish-record`.
-4. Preserve screenshot paths, errors, `needs_human` reasons, duration, account/device, media path, platform post id, permalink, and command family.
+4. Preserve screenshot paths, errors, failure reasons, duration, account/device, media path, platform post id, permalink, and command family.
 5. For Reddit permalink lookup, avoid marking app-level success as failed only because public lookup lacks a matching username.
 6. Hand successful records to Mobile Publish Data Recovery.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@11agents/cli",
-  "version": "0.1.25",
+  "version": "0.1.27",
   "description": "11agents local runtime and telemetry CLI",
   "type": "module",
   "bin": {

package/src/commands/mobile.js

@@ -1,4 +1,5 @@
 import { spawn } from 'node:child_process'
+import { createHmac } from 'node:crypto'
 import { constants } from 'node:fs'
 import { access, copyFile, mkdir, readdir, readFile, writeFile } from 'node:fs/promises'
 import os from 'node:os'
@@ -65,6 +66,7 @@ const COMMAND_ALIASES = new Map([
 const WRAPPER_FLAG_NAMES = new Set([
   'mobile-home',
   'task-id',
+  'task-title',
 ])
 
 const EMPTY_DATA_FILES = [
@@ -96,7 +98,9 @@ Examples:
 Runtime:
   Python/device-control code is bundled with @11agents/cli.
   Mutable runtime state lives in ~/.11agents/mobile by default.
-  Every command writes logs under ~/.11agents/mobile/runs/<task_id>/log.`)
+  Every command writes logs under ~/.11agents/mobile/runs/<task_id>/log.
+  Publish-success DingTalk notification is enabled by ELEVENAGENTS_PUBLISH_DINGTALK_WEBHOOK
+  and optional ELEVENAGENTS_PUBLISH_DINGTALK_SECRET.`)
 }
 
 function mobileHome(flags = {}, env = process.env) {
@@ -492,11 +496,19 @@ function buildRunContext(rawArgs, deps = {}) {
   const env = deps.env || process.env
   const home = mobileHome(flags, env)
   const taskId = taskIdFrom(flags, env)
+  const taskTitle = String(
+    flag(flags, 'task-title')
+      || env.ELEVENAGENTS_TASK_TITLE
+      || env.ELEVENAGENTS_TASK_NAME
+      || env.ELEVENAGENTS_ISSUE_TITLE
+      || taskId
+  ).trim()
   const runDir = path.join(home, 'runs', taskId)
   return {
     flags,
     home,
     taskId,
+    taskTitle,
     runDir,
     logPath: path.join(runDir, 'log'),
   }
@@ -526,6 +538,14 @@ async function dispatchPython(rawArgs, deps = {}) {
     exit_code: result.code,
     parsed,
   }, null, 2))
+  await notifyMobilePublishSuccess({
+    command,
+    context,
+    parsed,
+    env: deps.env || process.env,
+    logPath: context.logPath,
+    deps,
+  })
   if (result.stdout) process.stdout.write(result.stdout)
   if (result.stderr) process.stderr.write(result.stderr)
   if (result.code !== 0) process.exitCode = result.code
@@ -573,6 +593,124 @@ function parseJsonResult(text) {
   return rows.length ? rows : null
 }
 
+async function notifyMobilePublishSuccess({ command, context, parsed, env = process.env, logPath, deps = {} }) {
+  const config = dingTalkConfig(env)
+  if (!config.webhook || !Array.isArray(parsed)) return
+  const platform = platformFromMobileCommand(command)
+  if (!platform) return
+  for (const row of parsed) {
+    if (!isPublishedResult(row)) continue
+    const link = publishResultLink(row)
+    if (!link) continue
+    const content = buildPublishNotificationContent({
+      platform,
+      task: context.taskTitle || context.taskId,
+      deviceId: row.device_id || row.device || '',
+      link,
+    })
+    try {
+      const response = await (deps.sendDingTalkText || sendDingTalkText)({
+        ...config,
+        content,
+      })
+      await appendLog(logPath, `notification: ${content}\n`)
+      if (response) await appendLog(logPath, `notification_response: ${response}\n`)
+    } catch (error) {
+      await appendLog(logPath, `notification_error: ${errorMessage(error)}\n`)
+    }
+  }
+}
+
+function dingTalkConfig(env = process.env) {
+  return {
+    webhook: env.ELEVENAGENTS_PUBLISH_DINGTALK_WEBHOOK || env.ELEVENAGENTS_DINGTALK_WEBHOOK || '',
+    secret: env.ELEVENAGENTS_PUBLISH_DINGTALK_SECRET || env.ELEVENAGENTS_DINGTALK_SECRET || '',
+  }
+}
+
+function platformFromMobileCommand(command = '') {
+  const text = String(command || '').toLowerCase()
+  if (!text.startsWith('publish-')) return ''
+  if (text.includes('xiaohongshu')) return 'xiaohongshu'
+  if (text.includes('tiktok')) return 'tiktok'
+  if (text.includes('instagram')) return 'instagram'
+  if (text.includes('facebook')) return 'facebook'
+  if (text.includes('reddit')) return 'reddit'
+  if (text === 'publish-x' || text.includes('publish-x-')) return 'x'
+  return ''
+}
+
+function isPublishedResult(row) {
+  return row && typeof row === 'object' && row.status === 'published'
+}
+
+function publishResultLink(row) {
+  return String(
+    row.platform_permalink
+      || row.permalink
+      || row.url
+      || row.link
+      || ''
+  ).trim()
+}
+
+function buildPublishNotificationContent({ platform, task, deviceId = '', link }) {
+  const label = platformLabel(platform)
+  const taskText = String(task || 'task').trim()
+  const deviceText = deviceId ? `（设备 ${deviceId}）` : ''
+  return `${label} ${taskText}${deviceText}已发布，链接：${link}`
+}
+
+function platformLabel(platform) {
+  const labels = {
+    xiaohongshu: '小红书',
+    tiktok: 'TikTok',
+    instagram: 'Instagram',
+    facebook: 'Facebook',
+    reddit: 'Reddit',
+    x: 'X',
+  }
+  return labels[platform] || platform
+}
+
+async function sendDingTalkText({ webhook, secret = '', content }) {
+  const url = signedDingTalkUrl(webhook, secret)
+  const response = await fetch(url, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json;charset=utf-8' },
+    body: JSON.stringify({
+      msgtype: 'text',
+      text: { content },
+    }),
+  })
+  const text = await response.text()
+  if (!response.ok) throw new Error(`DingTalk webhook failed with HTTP ${response.status}: ${text}`)
+  try {
+    const parsed = JSON.parse(text)
+    if (parsed?.errcode) throw new Error(`DingTalk webhook failed: ${text}`)
+  } catch (error) {
+    if (error instanceof SyntaxError) return text
+    throw error
+  }
+  return text
+}
+
+function signedDingTalkUrl(webhook, secret = '', now = Date.now()) {
+  if (!secret) return webhook
+  const timestamp = String(now)
+  const sign = encodeURIComponent(
+    createHmac('sha256', secret)
+      .update(`${timestamp}\n${secret}`)
+      .digest('base64')
+  )
+  const sep = webhook.includes('?') ? '&' : '?'
+  return `${webhook}${sep}timestamp=${timestamp}&sign=${sign}`
+}
+
+function errorMessage(error) {
+  return error instanceof Error ? error.message : String(error)
+}
+
 function removeFlags(args, names) {
   const out = []
   for (let i = 0; i < args.length; i += 1) {
@@ -686,9 +824,14 @@ function restAndFlagsAfter(argv, positionalCount) {
 }
 
 export const mobileInternals = {
+  buildPublishNotificationContent,
   dataCollectCommand,
+  dingTalkConfig,
   mobileHome,
   parseRawArgs,
+  platformFromMobileCommand,
+  publishResultLink,
+  signedDingTalkUrl,
   stripWrapperFlags,
   taskIdFrom,
   venvPython,

package/src/commands/runtime.js

@@ -843,6 +843,18 @@ function agentNameForTask(task) {
   return task.agent?.name || task.agent?.id || task.agent_id || task.assignee_id || 'agent'
 }
 
+function taskTitleForEnv(task) {
+  return String(
+    task.title
+      || task.name
+      || task.issue?.title
+      || task.queue_event?.trigger_title
+      || task.queue_event?.trigger_summary
+      || task.id
+      || ''
+  )
+}
+
 function normalizeSkillBundle(skill = {}) {
   return {
     id: String(skill.id || ''),
@@ -1015,6 +1027,7 @@ async function prepareRuntimeTask(task, flags, deps, config) {
     ELEVENAGENTS_AGENT_RESULTS_DIR: agentResultsDir,
     ELEVENAGENTS_TASK_TMP: tmpDir,
     ELEVENAGENTS_TASK_ID: String(task.id || ''),
+    ELEVENAGENTS_TASK_TITLE: taskTitleForEnv(task),
     ...(projectToken ? { GTM_SWARM_TOKEN: projectToken } : {}),
   }
 
@@ -1061,6 +1074,21 @@ function buildCodexPrompt(task) {
     '',
     'MCP tool guidance:',
     '- If the task needs image generation, use the hosted 11agents MCP `image_generate` tool when available.',
+    '- Recommended image_generate defaults: model="nano", nano_model="preview", aspect="1:1", image_size="2K".',
+    '- nano_model alternatives: "preview", "pro", "flash". Use "preview" for reliable default quality, "pro" for harder prompts, and "flash" for faster drafts.',
+    '- aspect alternatives: "1:1", "16:9", "9:16", "4:3", "3:4". Choose the closest ratio to the final placement.',
+    '- image_size alternatives: "1K", "2K". For a specific output size, pass aspect and image_size; pass size only when the MCP tool/provider explicitly supports the requested exact dimensions.',
+    '- Use model="gpt" with model_id only when the task or current project notes confirm that GPT image generation is available for the hosted MCP key.',
+    '',
+    'Final task comment requirements:',
+    '> [!IMPORTANT]',
+    '> Default to a directly reviewable final comment. Do not only say the task is done.',
+    '- Every deliverable must appear in the final comment as a Markdown link to the file in agent_results_dir, for example `[final image](<absolute path>)` or `[report](<absolute path>)`.',
+    '- Put generated images, reports, datasets, and other final artifacts in agent_results_dir before linking them. The platform will upload allowed local links from agent_results_dir and replace them with reviewable asset URLs.',
+    '- Include a `## Deliverables` section first when files or generated media exist. Include a compact `## Review Notes` section after it with what changed and what to inspect.',
+    '- Use Markdown tables for comparisons, parameter lists, changelogs, data analysis, and multi-field entries.',
+    '- Use `##` and `###` headings, blank lines between paragraphs, bold keywords for important labels, and GitHub alert blocks for important notes or warnings.',
+    '- Use ordered lists for step-by-step workflows and unordered lists for parallel facts. Use fenced code blocks with a language tag, and use mermaid for flows or architecture when a diagram is useful.',
     '',
     'Task context:',
     compactJson({