Introduction

When developing agents with DSH, the common scenario is “session-based ephemeral use”: thoughts that pop up during the process or tasks mentioned on the fly are scattered once the session ends. Of course, opening a separate note-taking app is possible, but capturing, organizing, and producing requires manual back-and-forth. BigSeed’s approach is to delegate this to the agent: capturing life snippets and insights during daily use, accumulating them to build a user profile from these fragments, and ultimately generating novels, movie scripts, or autobiographies starring the user.

It is a community skill, not an official DeepSeek product. Below, I will introduce its positioning, features, installation, and precautions.

What is this

BigSeed is a DSH skill maintained by kobenfang. One-sentence positioning: Flash Memory Recording and Life Puzzle — capturing life snippets and insights, building a user profile from fragmented information, and generating novels, movie scripts, or autobiographies starring the user.

Key basics:

  • npm package name @kobenfang/dsh-bigseed, version 1.0.0, license MIT
  • The skill name in DSH is bigseed
  • Directory structure consists of only two parts: SKILL.md (skill definition, Agent Skills standard) and scripts/bigseed.py
  • Also published on ClawHub (OpenClaw platform), included in the community directory skillhub.cn

It follows the Agent Skills open standard (SKILL.md + YAML frontmatter), so the same skill can be used in frameworks like OpenClaw and Claude Code, which support this standard, in addition to DSH.

Core Features

The verified features are four:

  1. Flash Memory Recording: Capturing life snippets and insights.
  2. User Profile: Built step-by-step from fragmented information.
  3. Content Generation: Producing novels, movie scripts, or autobiographies starring the user.
  4. Cross-framework: Follows Agent Skills open standard, applicable to DSH, OpenClaw, Claude Code, and other frameworks.

The specific trigger methods for each function are not expanded in the README; detailed instructions point to the SKILL.md in the repository, which can be viewed directly after installation.

Installation and Activation

First, confirm two things: The environment requires Python 3; generating stories and profiles requires calling an LLM, so you need to configure the API for the corresponding model.

The installation commands are as follows; choose either method:

Method 1 (Recommended): Place in a user-level shared directory for use across projects:

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

Method 2: Place in the current project directory, effective only within the project:

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

After installation, DSH can be started to find bigseed in the skill list.

OpenClaw users can also use ClawHub:

npx clawhub install bigseed
# Search to confirm: npx clawhub search bigseed

Typical Usage

The reproducible path given in the README is the two installation commands above plus one confirmation step: start DSH, find bigseed in the skill list, and then use it according to the instructions in SKILL.md.

If you need to search for this skill in the directory, the keywords are: flash memory, life story, biography, autobiography, diary, journal, life story, biography, memory.

Suitable Scenarios and Precautions

Suitable for:

  • Users who want an agent to help them record thoughts and keep a diary long-term
  • Writers interested in “generating narrative works from personal fragments”
  • Developers who use multiple frameworks supporting Agent Skills, such as DSH, OpenClaw, and Claude Code, and want to reuse the installation

Three precautions:

  1. The skill runs with the permissions of the current DSH process. You should check the source code and license before installing any third-party skill. BigSeed is MIT; the source code consists only of SKILL.md and scripts/bigseed.py, so the check cost is low.
  2. Requires Python 3 and an available model API. Generating stories and profiles involves LLM calls, so configure it before enabling.
  3. This skill will build a user profile from the fragments you record. If the content is private, it is recommended to read SKILL.md and scripts/bigseed.py first to confirm the handling method before using it long-term.

Conclusion

To summarize: BigSeed turns flash memory recording, profile building, and story generation into a continuous flow. Once installed in DSH, it becomes an on-demand recording and creative skill. Refer to SKILL.md in the repository for specific functional details.

  • Community Directory Page: https://www.skillhub.cn/plugins/kobenfang/BigSeedSkill (Independent community site, no official affiliation with DeepSeek / Fenxiang)
  • GitHub Repository: https://github.com/kobenfang/BigSeedSkill