The performance of a Markdown knowledge base in search engines and generative engines often depends on how structured and interpretable its content is. Without automated tooling, performing SEO (Search Engine Optimization), GEO (Generative Engine Optimization), and AEO (Answer Engine Optimization) checks is tedious. DeepSeek Harness provides plugin capabilities, and dsh-geo, as one Bundle in it, standardizes this optimization workflow so that an Agent can directly audit and plan a knowledge base.

Plugin positioning

dsh-geo is a DeepSeek Harness Bundle maintained by winyh. It focuses on providing interpretable SEO, GEO, and AEO tools for Markdown knowledge bases.

This plugin is classified as a “networking tool” and uses the MIT License. It does not involve logging in, submitting forms, or batch publishing. Instead, it focuses on content auditing, metadata inspection, and content outline generation, helping developers transform product language into content that search engines and AI engines can understand, trust, and cite.

Core features

dsh-geo mainly handles content discoverability and retrievability. Its core capabilities include:

  1. Single-note audit: Performs SEO, GEO, and AEO readiness checks on a single Markdown note.
  2. Knowledge base scan: Scans an entire Vault to identify missing metadata, missing sources, broken knowledge structure, orphaned notes, and duplicate titles.
  3. Project report: Checks the configured root directory before scanning and generates a shareable project report.
  4. Content outline: Builds a content outline based on query intent, target audience, outline, question list, and source gaps.
  5. End-to-end workflow: Supports running a complete workflow directly from a public website URL, public account URL, Markdown export, or HTML snapshot.
  6. Keyword planning: Builds a qualitative keyword plan based on source signals, optional seed keywords, and Harness web search results (it does not make things up out of thin air).

Installation and enabling

Before installation, ensure the Node.js version meets the requirements:

node --version
# 需满足 ^22.19.0 || >=24.0.0

Use the following command to install. Make sure you install it in the Profile you actually start. For example, if you use dsh web, install it in the web profile; if you use dsh default, you must install it in the default profile, but installing it in default does not automatically enable it in web.

# 安装到 web profile
dsh plugin --profile web add github:winyh/dsh-geo

This plugin has been tested for compatibility with DeepSeek Harness 0.1.5-rc.2. The DSH peer dependency is locked to this tested version. If you use another release channel, you need to check compatibility again.

Typical usage

In DSH’s multi-plugin collaboration system, dsh-geo usually participates at the content layer. When starting a cross-plugin workflow, you can use the following interface:

# 接收上游任务或进行交接
geo_handoff_receive

If installed from GitHub source code, you need to use the prepare script to build the entry point. If you encounter a pnpm build block, you need to add the corresponding package key to allowBuilds in the profile’s pnpm-workspace.yaml and retry. For reproducibility, it is recommended to lock the reviewed Git commit.

Notes

  1. Permission scope: The plugin runs with the permissions of the current DSH process. It does not log in to websites, submit forms, or post in bulk.
  2. Private files: Private docs/ files are excluded from Git and packaged content.
  3. Profile logic: Installing to the default profile does not make the plugin active in the web profile; consistency must be maintained.
  4. Dependency versions: DSH peer packages are locked to the tested version. Do not mix packages from different release channels.

Summary

dsh-geo solves the problem of putting Markdown knowledge base SEO, GEO, and AEO tooling into practice. Through audits, scans, and content outline construction, it helps developers start from product facts and business goals to generate high-quality content assets that can be retrieved and cited by engines. For developers who need to build an interpretable knowledge base, this is a directly usable toolset.

For more details and source code, refer to the GitHub repository.