Introduction

In DSH (DeepSeek Harness), adding capabilities to agents usually involves writing scripts or integrating toolchains. However, work like product research is different: its core lies in analysis and structured output, where scripts are not very helpful. What really works is a set of well-defined prompts.

bigplan is an attempt to encapsulate such work into a skill. Below, I introduce its positioning, installation methods, and usage.

What is BigPlan

bigplan is an Agent Skill maintained by kobenfang. Its one-sentence positioning is: analyze market, technology, and supply chain based on product direction, and output three sets of product specification plans (high, medium, and low).

It follows the Agent Skills open standard (SKILL.md + YAML frontmatter), so it is not limited to DSH; frameworks that support this standard, such as OpenClaw and Claude Code, can also load it. The npm package name is @kobenfang/dsh-bigplan, version 1.0.0, and the license is MIT.

The repository structure is simple, with the core being a single skill definition file:

bigplan/
├── SKILL.md          # Skill definition (Agent Skills standard)

Core Features

The verified capabilities are as follows:

  • Analyze market, technology, and supply chain based on product direction
  • Output three sets of product specification plans (high, medium, and low)
  • Cover product research, competitor analysis, market analysis, and product planning
  • Purely prompt-driven, no script dependencies; recommended to use with internet search enabled

Note the last point: it itself does not include scripts, and the quality of analysis depends on the model’s capability and the prompts. If the runtime environment supports internet search, it is recommended to enable it; otherwise, market-related content will have to rely solely on the model’s existing knowledge.

Installation and Usage

First, clone the repository, then start DSH to verify. The README provides two installation methods.

Method 1 (Recommended): Place in the user-level shared directory for reuse across multiple projects.

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

Method 2: Place in the current project directory; it only takes effect for this project.

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

After the above steps, bigplan can be found in the skills list once DSH starts.

If you are using OpenClaw, this skill is also published on ClawHub (clawhub.ai) and can be installed directly:

npx clawhub install bigplan

For search verification:

npx clawhub search bigplan

Typical Usage

After installation, the skill is triggered by the framework on demand. Complete functionality descriptions are in the SKILL.md file in the repository. The search keywords provided in the README include:

Product research, market analysis, competitor analysis, product planning, product research, market analysis

In actual use, just describe the requirements around these keywords in the conversation. For example, ask the agent to conduct a complete research based on a specific product direction to obtain high, medium, and low specification plans. The output can be directly used as a starting point for project initiation discussions, with data manually supplemented and conclusions verified.

Applicable Scenarios and Notes

Who is it suitable for:

  • Independent developers or small teams who need to quickly produce product specification candidates before project initiation
  • People already using DSH or other frameworks that support Agent Skills, who want to solidify the research process into reusable skills

A few notes:

  1. bigplan is a pure prompt-based skill; output relies on the model and optional internet search. Conclusions require manual verification and cannot be used directly as a basis for decision-making.

  2. After installation, the skill runs with the permissions of the current DSH process. Before installing, it is recommended to read the repository source code first—the core is just the SKILL.md file, which has very low overhead—and confirm that the MIT license is compatible with your usage.

Conclusion

The value of bigplan lies in standardizing work like product research—which “appears to need no plugins”—into a single skill file: put it in ~/.agents to reuse across different projects; the output is fixed to high, medium, and low specification plans, making comparison and trade-offs easy. If you are exploring product directions, you can add it to your skill library.

  • Community directory page: https://www.skillhub.cn/plugins/kobenfang/BigPlan
  • GitHub repository: https://github.com/kobenfang/BigPlan
  • More skills by the same author: https://github.com/kobenfang/dsh-skills

It should be noted that skillhub.cn is a community-maintained independent directory site and has no official affiliation with DeepSeek / Hypersphere.