coverage-tracker-setup

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Coverage Tracker Setup

覆盖追踪器搭建指南

This sets up a simple keyword tracker that the
coverage-tracker
skill runs for you. Think of it like a Google Alert: it watches for new coverage of the keywords you care about and tells you when something real shows up.
It is deliberately a different job from setting up newsjack monitoring. A coverage tracker answers "did my keyword get real coverage?" Newsjack monitoring (
newsjack-monitor-setup
) answers a bigger question: "can my company jump into a broader news story?" If the user wants the second one, use that skill instead.
本指南将搭建一个由
coverage-tracker
skill为你运行的简易关键词追踪器。它类似Google Alert:会监控你关注的关键词的新覆盖内容,并在出现有效相关内容时通知你。
请注意,它与新闻劫持监控(newsjack monitoring)的定位完全不同。覆盖追踪器用于回答“我的关键词是否获得了有效曝光?”,而新闻劫持监控(
newsjack-monitor-setup
)则用于回答更宏观的问题:“我的公司能否切入某个热点新闻事件?”如果用户需要后者,请使用对应的skill。

Where you're running this

运行环境说明

Two situations, and they change what you can finish here:
  • You can save files and run commands (Claude Code, Codex, OpenClaw, Hermes, or a similar setup with shell, file, and network access). Here you can do the whole thing: save the tracker, run it once, and hand the recurring schedule to your agent. Before you decide you can't run the CLI, check whether
    newsjack
    is installed.
    It ships as a prebuilt, bundled binary — you do not need Go, a compiler, or any build/install step. Never look for a Go toolchain or tell the user they need a "Go environment" without checking first: run
    newsjack --version
    ; if that's not on
    PATH
    , try the bundled location
    ~/.newsjack/bin/newsjack --version
    and use that full path everywhere if it works. The bundled binary is almost always already installed — verify, don't assume it's missing.
  • You're in a plain chat window (Claude.ai, ChatGPT, Claude Cowork, or anything without file/command access). Here you can only draft. Write out the tracker config and the schedule prompt, then tell the user to move to a setup that can save and run things. Don't try to install or run anything, and don't tell the user it was saved or scheduled when it wasn't.
两种运行场景会影响操作流程:
  • 可保存文件并执行命令(如Claude Code、Codex、OpenClaw、Hermes等具备shell、文件和网络访问权限的环境):在此环境下可完成全部操作:保存追踪器配置、立即运行一次,并将周期性任务交由你的Agent调度。在判定无法运行CLI前,请先检查
    newsjack
    是否已安装
    。它以预编译的捆绑二进制包形式发布——你无需安装Go语言、编译器或执行任何构建/安装步骤。不要直接假设缺少Go环境,先执行
    newsjack --version
    验证;若该命令不在
    PATH
    中,尝试使用捆绑路径
    ~/.newsjack/bin/newsjack --version
    ,如果可用则全程使用该完整路径。捆绑二进制包几乎已预先安装,请先验证,不要主观判断缺失。
  • 纯聊天窗口(如Claude.ai、ChatGPT、Claude Cowork等无文件/命令访问权限的环境):在此环境下仅能完成草稿编写。写出追踪器配置和调度提示语,然后告知用户需要切换到可保存并运行文件的环境完成后续操作。不要尝试安装或运行任何程序,也不要谎称已完成保存或调度。

What to ask the user

需要向用户询问的信息

Only ask for what you don't already know:
  • a name for the tracker
  • the keywords to watch (as many as they want)
  • for each keyword, one short note on what it actually means: which company, product, or person should it point to? This is what later filters out wrong matches.
  • optionally, words or uses to ignore when a keyword is ambiguous (for example, a common word that also happens to be a brand name)
  • how often to check: every morning, twice a day, or hourly
Don't ask about company standing, spokespeople, competitors, proof assets, news feeds, target reporters, or story angles. All of that is for newsjacking, not for tracking coverage of your own keywords.
仅询问未知信息:
  • 追踪器的名称
  • 需要监控的关键词(数量不限)
  • 每个关键词对应的具体含义:指向哪家公司、产品或个人?这将用于后续过滤错误匹配
  • 可选:当关键词存在歧义时需要忽略的词汇或用法(例如,某个常用词同时也是品牌名)
  • 检查频率:每日早间、每日两次或每小时一次
无需询问公司现状、发言人、竞争对手、素材资源、新闻源、目标记者或报道角度。这些信息属于新闻劫持场景,而非关键词覆盖追踪场景。

Steps to set it up

搭建步骤

  1. Write the tracker config. This is a small file that the
    coverage-tracker
    skill reads back every time it runs, so keep its shape exactly as shown:
    json
    {
      "name": "Profound",
      "lookback_days": 2,
      "keywords": [
        {
          "keyword": "profound",
          "means": "Profound, the AI search analytics company.",
          "exclude_hints": ["generic adjective uses"]
        }
      ]
    }
    You can list as many keywords as you want. The
    means
    line for each one matters: make it specific (which company, product, or person) so the tracker can later throw out mentions of the wrong thing or generic uses of the word.
    If you're in a plain chat window, stop here. Hand the user this config plus the schedule prompt from step 3, and tell them to finish in a setup that can save and run things. Don't claim it was saved or scheduled.
  2. Save the config with this command (replace
    <slug>
    with a short name for the tracker):
    bash
    newsjack coverage init <slug> --config tracker.json
    Working inside a copy of the project source, use
    bin/newsjack
    from the project root instead. Only add
    --force
    if the user clearly wants to overwrite a tracker that already exists.
  3. Schedule the recurring check through your agent, not your computer's built-in scheduler. If your agent has a way to schedule prompts, schedule this one:
    text
    Use the coverage-tracker skill for <slug>.
    If your agent can't schedule things itself, tell the user exactly which prompt to set up on a schedule in Claude, Codex, Hermes, OpenClaw, or whatever they use. Do not set up system-level schedulers (cron, launchd, systemd timers, or similar) on the user's machine.
  4. Run it once right away if the user wanted a working setup, or if you're doing the whole setup end to end. Use the
    coverage-tracker
    skill for the new slug and tell the user what the first run found.
  1. 编写追踪器配置。这是
    coverage-tracker
    skill每次运行时都会读取的小型文件,请严格遵循如下格式:
    json
    {
      "name": "Profound",
      "lookback_days": 2,
      "keywords": [
        {
          "keyword": "profound",
          "means": "Profound, the AI search analytics company.",
          "exclude_hints": ["generic adjective uses"]
        }
      ]
    }
    你可以添加任意数量的关键词。每个关键词的
    means
    字段至关重要:请写清楚具体指向(哪家公司、产品或个人),以便追踪器后续排除错误提及或词汇的通用用法。
    若处于纯聊天窗口环境,到此为止。将该配置和步骤3的调度提示语交给用户,并告知其需切换到可保存并运行文件的环境完成后续操作。不要声称已完成保存或调度。
  2. 保存配置,执行如下命令(将
    <slug>
    替换为追踪器的短名称):
    bash
    newsjack coverage init <slug> --config tracker.json
    若在项目源码副本中操作,请使用项目根目录下的
    bin/newsjack
    。仅当用户明确需要覆盖已存在的追踪器时,才添加
    --force
    参数。
  3. 通过Agent而非系统内置调度器设置周期性检查。如果你的Agent支持调度提示语,请调度如下内容:
    text
    Use the coverage-tracker skill for <slug>.
    若你的Agent不支持自主调度,请明确告知用户需在Claude、Codex、Hermes、OpenClaw等工具中设置该提示语的调度任务。不要在用户机器上设置系统级调度器(如cron、launchd、systemd timers等)。
  4. 立即运行一次(如果用户需要立即可用的追踪器,或你正在完成完整的端到端搭建)。使用
    coverage-tracker
    skill运行新的slug,并告知用户首次运行的结果。

Updating a tracker you already have

更新已有的追踪器

First, look up where the config lives:
bash
newsjack coverage status <slug>
That shows a
config_path
. Open the config file there, make your edits, then save it again with:
bash
newsjack coverage init <slug> --config tracker.json --force
Only change the keywords or their meaning notes. The record of what's already been seen and alerted on is kept separately by
coverage-tracker
(in its own database) — don't edit that by hand.
首先,查询配置文件的位置:
bash
newsjack coverage status <slug>
该命令会显示
config_path
。打开该路径下的配置文件,进行编辑,然后执行如下命令保存更改:
bash
newsjack coverage init <slug> --config tracker.json --force
仅修改关键词或其含义说明。
coverage-tracker
会单独保存已监控和已通知的记录(存储在自身数据库中)——请勿手动编辑该部分内容。

What to tell the user when you're done

完成后的告知内容

Wrap up by giving them:
  • the tracker slug (its short name)
  • where the config file is saved
  • the schedule prompt and how often it'll run
  • what the first run found, if you ran it
结束时需向用户提供:
  • 追踪器的slug(短名称)
  • 配置文件的保存位置
  • 调度提示语及运行频率
  • 首次运行的结果(若已执行)