dsh-project-skill-paths
Run the following command in DeepSeek Harness:
dsh plugin install NeoRrrr/dsh-project-skill-paths
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install NeoRrrr/dsh-project-skill-paths in your terminal to install this plugin; full source: https://github.com/NeoRrrr/dsh-project-skill-paths
About this plugin
When juggling several projects in DeepSeek Harness, global and user-level Skill directories tend to accumulate and cross-contaminate—Skills meant for one project leak into another. dsh-project-skill-paths pushes Skill configuration into each Git project: it reads .dsh/skill-paths.yaml at the project root and adds the listed directories to that project's Skill discovery chain, so every project only loads the Skills it actually needs.
Through the DSH Desktop settings card you can add or remove any number of project-relative Skill roots; changes hot-reload without a restart. By default the plugin rejects absolute paths and relative paths that escape the project, keeping the configuration boundary explicit. Configured roots rank at 250—after built-in project roots (100/200) but before global and user-level roots (300+), so a same-name Skill resolves to the project-local copy first.
If you manage multiple DSH workspaces or team repositories and want Skill files to stay inside their own project rather than polluting a shared directory, this plugin is built for that. It reads no credentials, makes no outbound requests, and writes only the current project's .dsh/skill-paths.yaml.
Use Cases
- Preventing cross-project Skill contamination when running multiple DSH workspaces
- Isolating repository-specific Skill files within their own project
- Hot-reloading Skill directory changes without restarting DSH
Best For
- Users with multiple DSH workspaces or team repositories
- Developers who want project-local Skill files instead of a bloated global directory
- Engineers using DSH Desktop for daily coding workflows
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.