Skip to main content

How to use dsh-plugin-claude-bridge

In DeepSeek Harness, directly read and reuse Claude Code's memory, skills, and global instructions. No file migration needed—let both tools share the same memory and skills.

This article is auto-derived from indexed fields (wiki / faq / compatibility_json), not freshly AI-generated.

This article is derived from the plugin's already-indexed fields (wiki / faq / compatibility_json / readme), not freshly generated by AI. Source field is noted at the end of each section.

Quick start

dsh-plugin-claude-bridge

— source: plugin_wiki.wiki_content

Install & verify

dsh plugin --profile web add dsh-plugin-claude-bridge

Run the command above in your DSH Web Profile. Then enable the plugin in the plugin list.

— source: plugins.install

Key points

  • ~/.claude/projects/<project>/memory/*.md -- Injects memories as dynamic system prompt context
  • ~/.claude/skills/<name>/SKILL.md -- Adds skills to the available catalog
  • ~/.claude/CLAUDE.md -- Injects global instructions into system prompt
  • id: claude-bridge
    • '~/.agents/skills'

— source: plugin_wiki.readme_en (fallback readme_raw)

FAQ

What can this plugin do?

It reads the memory, skills, and global CLAUDE.md already stored by Claude Code (Anthropic's command-line programming assistant) and directly injects them into DeepSeek Harness's system prompt, allowing the two tools to share the same memory and skills without manually migrating files.

Do I need to copy or migrate files first?

No. The plugin reads the original files directly from ~/.claude/ at runtime, read-only without modification; if you don't want to use it, just uninstall the plugin and Claude Code's data remains unchanged.

Will it modify my content in Claude Code?

No. All reads in the source code use fs/promises readdir/readFile, there are no write operations and no symbolic links are created.

I only want to use memory, not skills - how do I disable it?

Set the corresponding switch to false in the plugin configuration, for example enableSkills: false; not writing it defaults to enabled.

What is the order of the three types of content in the system prompt?

The global CLAUDE.md is placed first with order=5, memory uses order=120, and the skills directory uses order=121; higher priority is given to items appearing earlier (see src/index.ts:74, 90, 106).

What if the project memory cannot be found?

The plugin encodes the current working directory into a project key using Claude Code's rules, for example C:\Users\foo becomes C--Users-foo; if the current directory has no corresponding projects subdirectory, the injected content will be empty.

How to add extra skills directories?

Use extraSkillDirs configuration to append directory paths, a common写法 is ['~/.agents/skills'], which will be merged with the default ~/.claude/skills for scanning.

— source: plugin_wiki.faq_json

Compatibility

  • DSH: 未声明（依赖宿主提供 @deepseek-ai/cordis 服务）
  • Node: ^22.19.0 || >=24.0.0

— source: plugin_wiki.compatibility_json

Pitfalls

Review the upstream repo before installing. This guide is auto-derived from indexed fields and may lag the latest release. If anything contradicts the official docs, treat the upstream source as authoritative.

— source: general rule