← Back to Skills Marketplace
yuewuya20180928

热点检索输出报文

by yuewuya20180928 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
43
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install redian-jiansuo-baochuwen
Description
定时从白名单企业(默认 OpenAI·Google·NVIDIA·Anthropic·DeepSeek·阿里巴巴·字节跳动·Tesla·SpaceX)抓取24小时内热点,用搜索工具 content摘要直接撰写中文短文(默认400 字),发送纯文字(默认3 条、无图)到推送渠道。模型、搜索、白名单、字数、推送渠道、...
README (SKILL.md)

热点检索输出报文

定时向推送渠道发送多条科技/AI热点纯文字短文。所有关键参数可配置。

适用场景

  • 用户希望定时收到 AI / 科技 / 互联网行业资讯
  • 推送渠道任意(支持微信 iLink AI、Telegram、Slack、Email 等)
  • 内容限定在白名单企业内
  • 要求事实可查、来源可溯、不发图、不幻觉

配置变量(全部可改)

修改这些变量即可改变任务行为,无需改 prompt / cron job 主体。

变量 默认值 说明
MODEL minimax/MiniMax-M2.7 生成模型。也可选 M3、Claude Sonnet/Opus 等
SEARCH_PROVIDER tavily 搜索后端。也可选 brave / searxng
WHITELIST OpenAI·Google·NVIDIA·Anthropic·DeepSeek·阿里巴巴·字节跳动·Tesla·SpaceX 话题来源限定(用 · 分隔)
ARTICLE_COUNT 3 每次推送条数(1-5 推荐)
WORDS_PER_ARTICLE 400 每条字数(推荐 300-600)
SEND_IMAGE false 是否发图。本机抓图不可靠,强烈建议保持 false
SCHEDULE_EXPR 15 23 * * * cron 表达式(默认每天 23:15)
STAGGER_MS 2700000 抖动毫秒数(45min)。设为 0 则固定时间触发
SCHEDULE_TZ Asia/Shanghai 时区
DELIVERY_CHANNEL openclaw-weixin 推送渠道(如 openclaw-weixin / telegram / email)
RECIPIENT 收到人 ID,格式依 channel 而定 接收人 ID
DELIVERY_ACCOUNT_ID 已配对的 bot 实例 ID 推送渠道对应的已配对账号
SENSITIVE_WORDS 武器/生物/政治/中国政策 触发内容过滤的话题关键词,遇到则换话题
TIMEOUT_SECONDS 300 任务超时

完整变量定义references/variables.md。修改后用 envsubst 或手动替换 ${VAR} 即可重新生成 prompt。

反幻觉 / 反内容过滤硬约束

- 只准用 ${SEARCH_PROVIDER} content 摘要里写明的事实,禁止脑补数字、人名、日期
- 不发图、不调用 image_generate(SEND_IMAGE=false 时强制)
- 避免触发内容过滤:不写 "${SENSITIVE_WORDS}" 等敏感词;
  如选中的话题包含敏感词,换个白名单企业的话题
- 话题必须来源于白名单(${WHITELIST}),其他公司一律不发
- 每篇必须有可验证的来源 URL(${SEARCH_PROVIDER} 返回的 source URL 字段)
- 如果凑不到 ${ARTICLE_COUNT} 条真实热点,宁可少发也不要凑数或编造

不需要做的事

  • 不要调用 web_fetch —— 本机抓 Bloomberg / Reuters / Yahoo 全部超时或 403
  • 不要下载图片 —— ${SEARCH_PROVIDER} 不返回稳定 image_url,且 SEND_IMAGE=false
  • 不要使用 Unsplash / 通用图库
  • 不要用 image_generate 编造配图(除非 SEND_IMAGE=true 且明确需要占位图)

操作步骤(agent 视角)

  1. web_search 搜索白名单(${WHITELIST})过去 24h 内的最新动态(中英文各一次,并行)
  2. 筛选出最有爆点的 ${ARTICLE_COUNT} 条热点(最好来自不同公司)
  3. 仅用 ${SEARCH_PROVIDER} 返回的 content 字段(200-500 字摘要)作为正文基础
  4. 每条撰写 ${WORDS_PER_ARTICLE} 字左右中文短文,文末标注「来源:[媒体名] [URL]」
  5. message 工具依次发到 ${RECIPIENT}
  6. 任务结束简短报告:发了 X 条、每条的企业归属、每条来源 URL、字数

重建 / 复用本任务(按变量配置)

1. 确认基础设施

# 推送 channel 已配对
openclaw channels list

# 搜索工具已配(${SEARCH_PROVIDER} 对应的 API key 已设置)
grep -E "TAVILY_API_KEY|BRAVE_API_KEY" ~/.openclaw/openclaw.json ~/.openclaw/secrets/*.env

# 模型已在 agents.defaults.models 列出
grep "models" ~/.openclaw/openclaw.json

2. 渲染 prompt 模板

# 用 envsubst 把 ${VAR} 替换为环境变量或默认值
envsubst \x3C references/prompt.md > /tmp/prompt-rendered.md

3. 创建 cron job

openclaw cron add \
  --name "daily-tech-news" \
  --schedule "${SCHEDULE_EXPR}" \
  --tz "${SCHEDULE_TZ}" \
  --stagger ${STAGGER_MS} \
  --model "${MODEL}" \
  --agent main \
  --target isolated \
  --timeout ${TIMEOUT_SECONDS} \
  --prompt-file /tmp/prompt-rendered.md \
  --delivery-mode announce \
  --delivery-channel "${DELIVERY_CHANNEL}" \
  --delivery-to "${RECIPIENT}" \
  --delivery-account "${DELIVERY_ACCOUNT_ID}"

4. 验证

openclaw cron run --id \x3Cjob-id>  # 手动触发
openclaw cron list               # 看状态
openclaw cron runs --id \x3Cjob-id>  # 看历史

配置切换示例

换模型为 M3

MODEL=minimax/MiniMax-M3 openclaw cron update --id \x3Cjob-id> \
  --patch '{"payload": {"model": "minimax/MiniMax-M3"}}'

换搜索工具为 Brave

# 1. 配 Brave API key
# 2. 改 openclaw.json: tools.web.search.provider = "brave"
# 3. 改 prompt 里的 "tavily" → "brave"
SEARCH_PROVIDER=brave openclaw cron update --id \x3Cjob-id> \
  --patch "$(cat \x3C(sed 's/tavily/brave/g' prompt.md))"

换字数

WORDS_PER_ARTICLE=500 envsubst \x3C references/prompt.md > prompt-500.md
openclaw cron update --id \x3Cjob-id> --prompt-file prompt-500.md

加新白名单企业

WHITELIST="OpenAI·Google·NVIDIA·Anthropic·DeepSeek·阿里巴巴·字节跳动·Tesla·SpaceX·Meta"
envsubst \x3C references/prompt.md > prompt-meta.md
openclaw cron update --id \x3Cjob-id> --prompt-file prompt-meta.md

切到 Telegram

DELIVERY_CHANNEL=telegram \
RECIPIENT=\x3Ctelegram-chat-id> \
DELIVERY_ACCOUNT_ID=\x3Ctelegram-bot-id> \
  envsubst \x3C references/prompt.md > prompt-tg.md
openclaw cron update --id \x3Cjob-id> --prompt-file prompt-tg.md
# 同时改 delivery 段

已知坑

1. iLink AI 沙盒化 / 14 分钟配额窗口

症状:cron 报 delivered:true,plugin 返回 messageId用户实际收不到任何消息

根因:iLink AI 后端对 sandbox/test bot 限制严格,配对后约 14 分钟内可正常投递,之后 sendMessage 端点进入"兼容式 stub"模式。

诊断命令(直接 curl 打 iLink AI):

# 获取 token
TOKEN=$(cat ~/.openclaw/openclaw-weixin/accounts/\x3Cbot-id>.json | python3 -c "import json,sys;print(json.load(sys.stdin)['token'])")

# Test: sendMessage
curl -s -w "\
[HTTP %{http_code}]" -X POST \
  -H "Content-Type: application/json" \
  -H "AuthorizationType: ilink_bot_token" \
  -H "X-WECHAT-UIN: dGVzdA==" \
  -H "Authorization: Bearer $TOKEN" \
  -d '{"msg":{"to_user_id":"\x3Copenid>@im.wechat","client_id":"cli_test","message_type":1,"message_state":2,"item_list":[{"type":1,"text_item":{"text":"test"}}]},"base_info":{"channel_version":"0.6.0","bot_agent":"openclaw/0.6.0"}}' \
  https://ilinkai.weixin.qq.com/ilink/bot/sendmessage
# 正常: {"errcode":0, ...}
# 沙盒化: {} + HTTP200

# Test: getConfig
curl -s -w "\
[HTTP %{http_code}]" -X POST \
  -H "Content-Type: application/json" \
  -H "AuthorizationType: ilink_bot_token" \
  -H "X-WECHAT-UIN: dGVzdA==" \
  -H "Authorization: Bearer $TOKEN" \
  -d '{"ilink_user_id":"\x3Cbot-id>@im.bot:\x3Ctoken-suffix>","base_info":{"channel_version":"0.6.0","bot_agent":"openclaw/0.6.0"}}' \
  https://ilinkai.weixin.qq.com/ilink/bot/getconfig
# 正常: {"ret":0, ...}
# 沙盒化: {"ret":-4,"errmsg":"GetTypingTicket rpc failed"} + HTTP200

临时缓解:重新配对拿新 token。 长期方案:登录 iLink AI 后台查看 bot 健康/配额,或换 channel( Telegram / Slack / Email)。

2. Cron delivery.accountId 指向已失效 bot

症状:cron 跑完报 delivered:true,但消息丢失。 原因:bot token 过期或被吊销,但 cron job 的 delivery.accountId 没同步更新。 修复:重配对后立即同步更新 cron job。

3. Web fetch 全失败

症状:Bloomberg / Reuters / Yahoo 等返回 30s 超时或 403。 修复:不再调用 web_fetch,只用 ${SEARCH_PROVIDER} content 字段。

4. 内容过滤 output new_sensitive

症状:模型输出截断,stopReason: error,message tool 调用被吞掉。 修复:SENSITIVE_WORDS 列表触发后主动换话题,不要写"AI"以外的元话题(行业政策/监管/中美关系)。

5. 旧 bot token 过期

症状:savedAt 超过 8 天的 token 在 iLink AI 后端被静默吊销。 修复:定期重配对。

重新配对微信 bot

# 1. 停 gateway(用 nohup 脱钩避免杀掉 exec session)
nohup systemctl --user stop openclaw-gateway > /dev/null 2>&1 &

# 2. 删旧 token 文件
rm ~/.openclaw/openclaw-weixin/accounts/\x3Cold-bot-id>.json
rm ~/.openclaw/openclaw-weixin/accounts/\x3Cold-bot-id>.sync.json
rm ~/.openclaw/openclaw-weixin/accounts/\x3Cold-bot-id>.context-tokens.json
echo '[]' > ~/.openclaw/openclaw-weixin/accounts.json

# 3. 启 gateway
systemctl --user start openclaw-gateway

# 4. PTY 模式跑 login 拿 QR
script -qfc "openclaw channels login --channel openclaw-weixin" /tmp/weixin-login.log

# 5. 手机微信扫 QR → 同意授权
# 6. 新 bot id 自动写入

# 7. 同步更新所有相关 cron job 的 delivery.accountId
# 8. 用 curl 验证新 bot 是否被沙盒化(见上文)

相关文件

  • references/variables.md — 完整变量定义与默认值
  • references/prompt.md — 完整 prompt 模板(含 ${VAR} 占位符)
  • references/cron-config.md — cron job 字段定义与默认值
Usage Guidance
Install only if you understand that it creates a recurring job that sends generated news summaries to external messaging services. Before enabling it, set your own recipient and account ID, remove or replace the hard-coded fallback WeChat recipient, avoid printing secrets from local config files, and back up bot account files before following any reset instructions.
Capability Tags
requires-sensitive-credentials
Capability Assessment
Purpose & Capability
The core purpose is coherent: it searches recent tech news, writes Chinese summaries, and sends them on a schedule to a configured delivery channel.
Instruction Scope
Outbound delivery is generally disclosed, but the prompt reference includes a concrete fallback WeChat recipient ID while other docs describe the recipient as user-supplied, creating a real risk of misdirected scheduled messages.
Install Mechanism
The package contains markdown instructions only; there are no executable scripts, hidden installers, or automatic install-time actions in the artifact.
Credentials
The skill asks users to inspect OpenClaw config and secret files, extract bot tokens, send direct authenticated curl requests, and modify cron delivery settings; these are related to the task but expose sensitive operational state.
Persistence & Privilege
Scheduled cron persistence is explicit and purpose-aligned, but the troubleshooting section uses raw rm commands against bot token/session files and rewrites an accounts file without backup or confirmation guardrails.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install redian-jiansuo-baochuwen
  3. After installation, invoke the skill by name or use /redian-jiansuo-baochuwen
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of 热点检索输出报文 (redian-jiansuo-baochuwen): - Scheduled delivery of recent tech/AI news summaries in pure text (default: 3 short articles, 400 characters each, from whitelisted companies only). - All key parameters are configurable, including model, search provider, whitelist, word count, delivery channel, recipients, and schedule. - Strict anti-hallucination and content filtering rules; only facts with verifiable sources, no images, no outside companies/topics, and sensitive words avoidance. - Flexible for various channels: supports WeChat (iLink AI), Telegram, Slack, Email, etc. - Includes detailed configuration guide, troubleshooting section, and operational steps for environment setup and maintenance.
Metadata
Slug redian-jiansuo-baochuwen
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is 热点检索输出报文?

定时从白名单企业(默认 OpenAI·Google·NVIDIA·Anthropic·DeepSeek·阿里巴巴·字节跳动·Tesla·SpaceX)抓取24小时内热点,用搜索工具 content摘要直接撰写中文短文(默认400 字),发送纯文字(默认3 条、无图)到推送渠道。模型、搜索、白名单、字数、推送渠道、... It is an AI Agent Skill for Claude Code / OpenClaw, with 43 downloads so far.

How do I install 热点检索输出报文?

Run "/install redian-jiansuo-baochuwen" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is 热点检索输出报文 free?

Yes, 热点检索输出报文 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does 热点检索输出报文 support?

热点检索输出报文 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created 热点检索输出报文?

It is built and maintained by yuewuya20180928 (@yuewuya20180928); the current version is v1.0.0.

💬 Comments