dsh-skill-mcp-bridge
Run the following command in DeepSeek Harness:
dsh plugin install cstarc/dsh-skill-mcp-bridge
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install cstarc/dsh-skill-mcp-bridge in DeepSeek Harness to install this plugin; the source repository is at https://github.com/cstarc/dsh-skill-mcp-bridge
About this plugin
Working with DeepSeek Harness, teams often find their skill definitions and MCP server declarations scattered across .claude, .agents, and .trae directories inherited from other toolchains. Re-declaring each one by hand in a new dsh project is tedious and prone to mistakes. dsh-skill-mcp-bridge was built for exactly this bridging scenario: after a single install it automatically discovers and imports every declared skill and MCP server, eliminating duplicated configuration.
Its core capabilities span three areas. Skill import scans .claude/skills, .agents/skills, .trae/skills, plus .claude/agents and .claude/commands, and parses frontmatter fields such as name, description, when_to_use, and disable-model-invocation so that both model-triggered and user-invoked skills behave correctly. MCP bridging merges mcpServers entries from .trae/mcp.json, .mcp.json, and .claude/settings.json in priority order, supporting both HTTP/SSE and stdio transports, and registers them as native dsh tools. The control panel offers dual views (settings page and tool view) with per-source toggles, rescan, and reconnection; a 4-second file-fingerprint poller hot-reloads configuration changes, and all toggle state is persisted locally across restarts.
It is aimed at developers who juggle multiple agent toolchains (Claude Code, Trae, custom agents) and want dsh as the unified runtime, as well as any engineer whose workspace already declares MCP servers and would rather not re-declare them inside dsh. No extra approval workflow is needed after installation.
Use Cases
- Unify scattered skills and MCP declarations from Claude Code, Trae, and custom agents into dsh
- Reuse existing .mcp.json or .claude/settings.json server definitions without re-declaring inside dsh
- Hot-reload skill or MCP configuration changes without restarting the dsh process
Best For
- Developers switching between multiple agent toolchains who want dsh as the unified runtime
- Engineers whose workspace already declares MCP servers or skills and would rather not reconfigure
- dsh users who need per-source GUI toggles to control what gets imported
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.