AI Agent Hub
Back to plugins
🎮

dsh-skill-switch

Fun & Dress-up Updated 2026.08.24

Run the following command in DeepSeek Harness:

dsh plugin install hzthzt/dsh-skill-switch

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install hzthzt/dsh-skill-switch in DeepSeek Harness to install this plugin; the full source is available at https://github.com/hzthzt/dsh-skill-switch .

About this plugin

When you maintain a large library of user-global Skills in DeepSeek Harness Web, you face a binary choice: expose every Skill at once, or hand-craft directory Junctions with no unified view and no safety net. dsh-skill-switch collapses that workflow into a single central directory. The plugin scans it one level deep, treats the frontmatter name in each SKILL.md as the canonical identity, and selectively maps chosen Skills into $DSH_HOME/skills via Windows directory Junctions you can enable, disable, or prune at will. It only creates and removes Junctions—source Skills are never installed, updated, edited, or deleted.

A five-state machine (available, enabled, conflict, invalid, broken) gives every anomaly a clear signal instead of a silent failure. All scans and mutations share a single serial queue; the ownership manifest is written atomically; and removal double-checks the Junction identity and destination before pruning, so an external tool claiming the same name can never trigger an accidental delete. On non-Windows hosts the plugin returns unsupported-platform and touches no filesystem at all.

If you use DSH Web daily, keep a large Skill library, and want to activate only the subset relevant to the current project or session—with every action auditable, reversible, and source-safe—this plugin was built for that workflow.

Use Cases

  • Activate only the Skills relevant to the current project from a large global library
  • Swap between different Skill sets when switching projects without cross-interference
  • Diagnose conflicts, broken links, or missing source directories with a clear state model

Best For

  • Power users who maintain a large library of DSH Web Skills
  • Teams that need project-level Skill isolation and clean handoff
  • Windows desktop users who value auditable, reversible Skill management