dsh-skills-anywhere
Run the following command in DeepSeek Harness:
dsh plugin install noteflowai/dsh-skills-anywhere
Paste the following prompt into your AI chat to install this plugin:
Install it in DeepSeek Harness by running dsh plugin install noteflowai/dsh-skills-anywhere; the full source code is available at https://github.com/noteflowai/dsh-skills-anywhere
About this plugin
When juggling multiple agent tools such as Claude Code, Codex, Cursor, or Gemini CLI, SKILL.md files end up scattered across project directories, user-level folders, and plugin caches with no unified discovery layer. Reusing the same set of skills across tools or pulling a shared team Git repository into daily workflows quickly turns into manual directory management and file copying that does not scale. dsh-skills-anywhere is a DeepSeek Harness plugin built to solve exactly that: it scans local directories, Claude Code plugin marketplaces, and configured Git sources into a single ranked catalog, then exposes the result through the built-in dsh skill tool or a standalone stdio MCP server so both the model and any MCP client can discover and load skills on demand.
Core capabilities revolve around a unified catalog, controllable loading, and auditable delivery. The catalog budget (default 50 entries) limits how many skill summaries enter the model session; the remainder stay reachable via find_skills keyword search and open_skill exact-name loading. Duplicates are deduplicated by symlink and byte identity, missing frontmatter fields are repaired in lenient mode with each fix logged, and naming conflicts are auto-renamed. Git sources maintain a local checkout with a lock file recording the resolved commit, synchronize on an interval in the background, and fall back to the existing cache on failure. MCP open_skill returns a SHA-256 receipt that callers can verify against expected_sha256 to reject a changed SKILL.md before instructions are delivered. The bundle subcommand generates a directory manifest for diffing added, removed, or modified resources. An interactive browser playground lets you drop in your own SKILL.md to compare strict versus lenient parsing, inspect repairs, and download a check report without installing a model or an API key.
The plugin is aimed at developers maintaining multiple agent toolchains who want a shared, versioned skill library, as well as security and compliance teams that need an auditable delivery chain for agent instructions. As long as your DeepSeek Harness profile mounts the dsh-skill runtime and your Node.js version is 22.19 or 24 or above, you can slot it into existing workflows without restructuring the skill directories of any other client.
Screenshots
Use Cases
- Reuse the same Agent Skills across Claude Code, Codex, Cursor, and other tools
- Share a team skill repository via Git with pinned commits and lock files
- Verify SKILL.md SHA-256 before loading instructions into a model session
Best For
- Developers maintaining multiple agent toolchains simultaneously
- Security and compliance teams auditing skill delivery pipelines
- Engineering leads treating shared skill libraries as team assets
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.