Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/jackwener/opencli/smart-searchnpx skills add jackwener/OpenCLI --skill smart-searchgit clone --depth 1 https://github.com/jackwener/OpenCLIWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00072 | $0.01798 |
| Opus 5 | $0.00036 | $0.00899 |
| Sonnet 5 | $0.00014 | $0.00360 |
| Haiku 4.5 | $0.00007 | $0.00180 |
Grade A, and why
smart-search scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
8 near-identical copies found in the catalogue:
- smart-search — 100% identical, 0 lines differ
- qiaomu-smart-search — 94% identical, 6 lines differ
- ldm-ai-to-human-zh — 89% identical, 682 lines differ
- humanize-ai — 89% identical, 393 lines differ
- novel-anti-detection — 89% identical, 240 lines differ
- ai-business-search-zh — 89% identical, 193 lines differ
- smart-search — 88% identical, 27 lines differ
- policy-monitor — 88% identical, 328 lines differ
How it starts
The opening of the file, as written. The whole thing — 157 lines — stays where its author put it; the contents beside it link to each section on GitHub.
智能搜索路由器
根据话题和场景,将查询路由到最佳的 opencli 搜索源。此 skill 的核心目标不是记忆命令,而是先定位数据源,再让 Agent 通过 opencli 自己读取实时帮助,避免文档漂移。
强制预检
每次使用前,必须先做下面两步:
- 运行
opencli list -f yaml - 用 live registry 确认候选站点是否存在,并检查
strategy、browser、domain
选定站点后,必须再做下面两步:
- 运行
opencli <site> -h查看该站点有哪些子命令 - 若已锁定某个子命令,再运行
opencli <site> <command> -h查看参数、输出列、策略
不要在 skill 文档里硬编码参数或假设命令签名;以 opencli ... -h 的实时输出为准。
主路由规则
只使用这一条规则,不再维护多套优先级:
- 当用户明确指定网站、平台或数据源时,直接使用对应网站。
- 当用户没有指定网站时,优先只选择一个 AI 源:
grok、doubao、gemini三选一。 - 当 AI 返回内容不足、缺少原始数据、需要权威佐证或需要垂直结果时,再补充 1-2 个专用源。
单题预算与频率限制
把“单个用户问题”理解为同一意图链路下的一次问题求解;同一轮追问、澄清、补充条件,若核心问题未变,仍算同一题。
先建立一份站点调用台账。每次真正执行搜索命令后,立刻更新:
sitequerycountstatus
计数规则:
opencli list -f yaml、opencli <site> -h、opencli <site> <command> -h属于预检与帮助,不计入搜索次数- 一次真正的
opencli <site> ...搜索/查询执行,计为该站点 1 次调用 - 同站点因为报错、超时、验证码、反爬、登录态异常而失败,也算 1 次调用,不要无限重试
频率上限:
- AI 站点硬限制:同一题内,每个 AI 站点最多调用 1 次
- 默认策略仍然是只选 1 个 AI 站点,不要把多个 AI 站点串成常规流程
- 只有当用户明确要求比较多个 AI 站点时,才可以额外调用其他 AI 站点;但每个被点名的 AI 站点仍然最多 1 次
- 非 AI 站点默认最多调用 2 次
- 非 AI 站点第 2 次调用必须有明确理由,例如第一次结果过宽,需要加时间、地区、类别、排序或关键词限定
- 非 AI 站点不要进行第 3 次调用;若信息仍不足,停止扩搜并明确说明缺口
触发限频后的处理:
- 记录:「已跳过: 达到频率上限」
- 优先改用其他同类站点
- 若没有合适替代源,则直接基于已收集信息回答,并说明覆盖范围与缺口
查询结束汇报
每次查询结束后,回答末尾必须追加一段简短的“搜索摘要”,至少包含下面三项:
- 使用了什么网站搜索
- 每个网站搜了什么词
- 每个网站搜了几次
如果有被限频跳过的站点,也要明确写出。
建议使用下面的固定格式:
搜索摘要
- 网站:<site1> | 查询词:<term1> | 次数:<n>
- 网站:<site2> | 查询词:<term2>;<term3> | 次数:<n>
- 已跳过:<site3>,原因:达到频率上限
AI 源选择
grok适合实时讨论、英文互联网舆论、Twitter/X 语境、热点追踪。doubao适合中文语境、字节抖音生态、生活方式内容、中文热点与泛中文问答。gemini适合全球网页、英文资料、通用信息检索、背景综述。
如果用户没有指定网站,默认先判断语言和语境,再从这三个里只选一个。
一旦某个 AI 站点已经执行过一次真实查询,就不要在同一题里改写关键词后再次调用该 AI 站点。若答案不足,优先补专用源,不要反复追打同一个 AI 站点。
AI 查询词建议
当使用 AI 源时,不要只丢一个过短关键词。优先构造成“主题 + 目标 + 限定条件”的查询。
- 主题 用户真正要查的对象、事件、产品、人物、公司、技术名词。
- 目标 想要什么结果,例如总结、对比、原因、趋势、推荐、原始线索。
- 限定条件 语言、地区、时间范围、平台范围、受众、价格带、岗位地点、是否要引用原始来源。
优先使用下面这种表达方式:
<主题> + <你要回答的问题><主题> + <时间范围/地区/语言><主题> + <平台或来源范围><主题> + <输出要求>
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 157 lines · 72 tokens per session scan A 0aa29dfe8c17
smart-search is a skill published in the GitHub repository jackwener/OpenCLI (28,734 stars, last pushed yesterday), licensed Apache-2.0. It adds 72 tokens to every session and 1,798 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
x402
Set up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key. Two setups it works out up front — "just use it" (set up a wallet so you or Claude Code can run cloud browser tasks paid from the wallet — Claude writes and runs throwaway scripts, nothing…
browser-use
Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work.
remote-browser
Controls an isolated Browser Use Cloud browser from a sandboxed machine with the current Browser Use CLI.
interceptor-research
Deep web-research methodology for the interceptor browser surface — investigate a topic the way researchers, intelligence analysts, investigative journalists, private investigators, and OSINT operators do, not 'read 3 links and summarize.' Use when the task is research, deep research, investigate, go deeper, be…
interceptor-browser
Drive a signed-in Chrome / Brave / Safari session via the interceptor CLI: open/read pages, click, type, inspect DOM/text/network, automate rich browser editors and scene graphs, capture WebSocket/Beacon/BroadcastChannel traffic, record/replay flows, take VLM-budgeted screenshots, compare pages, and route to specific…
interceptor-macos
Drive native macOS apps via interceptor macos : AX trees, background click/type/keys/drag/scroll, occluded or minimized window capture, browser chrome, URL bars, OS dialogs, Apple Events, trusted OS input, monitor/replay, overlays, vision, speech, NLP, JSC host utilities, Electron app web-content control via…