Preface

When building agent applications, we often encounter requirements such as: letting an agent continuously monitor certain information—product price drops, celebrity updates, industry news, custom metrics—and actively report them once changes occur. Relying on manual repeated queries is unrealistic, and writing a custom script from scratch with a scheduled task and notification function involves handling details like scheduling and multi-channel pushing.

bigfocus implements this as a DSH skill: once installed, it possesses the ability for scheduled pushing and multi-channel segmented pushing, automatically reporting when information changes. Below is an introduction to its positioning, features, and installation methods.

What is bigfocus

bigfocus is a DSH skill (Agent Skill) maintained by kobenfang. In one sentence, it is designed to track product prices/celebrity updates/industry information/custom metrics and automatically report when changes occur.

It follows the Agent Skills Open Standard (SKILL.md + YAML frontmatter), so in addition to DeepSeek Harness (dsh), it can also be used in frameworks that support this standard, such as OpenClaw and Claude Code. The code is released under an MIT license (noted in the License section of the README and package.json), the package name is @kobenfang/dsh-bigfocus, and the version is 1.0.0.

Core Features

According to the repository description, bigfocus provides the following capabilities:

  • Track product prices and automatically report when they change
  • Track celebrity updates
  • Track industry information
  • Track custom metrics
  • Scheduled push (based on cron scheduled tasks)
  • Multi-channel segmented push capabilities

The first few items cover common information tracking scenarios, while “custom metrics” allows you to define the objects you want to monitor yourself. The scheduling and pushing parts rely on cron. The repository provides relevant scripts and templates under the references/ directory, which will be used in the next section.

Installation and Activation

The environment dependencies are Python 3 and cron support. There are two ways to install.

Method 1 (Recommended): Place it in the user-level shared directory so multiple projects can share it:

mkdir -p ~/.agents
git clone https://github.com/kobenfang/bigfocus.git ~/.agents/bigfocus

Method 2: Place it in the current project directory:

mkdir -p .agents
cp -r bigfocus .agents/

After installation is complete, start dsh to find bigfocus in the skill list.

The repository directory structure is as follows:

bigfocus/
├── SKILL.md          # Skill definition (Agent Skills standard)
├── references/cron-install-shell.sh
├── references/cron-templates.json
├── scripts/bigfocus.py

When configuring scheduled tasks, you can refer to references/cron-install-shell.sh and references/cron-templates.json. For complete feature descriptions, see the SKILL.md in the repository.

OpenClaw users can also install from ClawHub:

npx clawhub install bigfocus

You need to confirm the search command first:

npx clawhub search bigfocus

Suitable Scenarios and Notes

Suitable scenarios: agents need to monitor certain types of information for a long time and receive notifications when they change, such as monitoring product price drops, tracking updates for specific people or industries, or monitoring custom metrics.

There are a few notes before installation:

  1. bigfocus depends on Python 3 and cron; confirm that your machine meets these two conditions before running.
  2. The plugin runs with the permissions of the current dsh process; it is recommended to read the repository source code (especially scripts/bigfocus.py) before installing to ensure the behavior meets your expectations.
  3. The license is MIT, so you may reference or modify it yourself after confirming the terms.

Conclusion

bigfocus solves a specific problem: handing over the “repeated query → change detection → multi-channel notification” chain to cron scheduled tasks and dsh skills, saving the repetitive labor of setting up scheduling and pushing. If your agent needs this kind of tracking capability, you can install and try it.