DeepSeek Harness plugin

dsh-skill-manager-thomaske

Manage DeepSeek Harness skills: list filesystem skills and toggle them on/off (SKILL.md <-> SKILL.md.disabled), from the agent or via the Settings UI (bundle-included).

Jump to install

Source facts

Repository
Thomas-key/dsh-skill-manager
Latest update
Aug 14, 2026
Category
Workflow & Automation
GitHub stars
0
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/Thomas-key/dsh-skill-manager
Plugin: dsh-skill-manager-thomaske
Author: Thomas-key

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer3 files
README.mdSource · read only

dsh-skill-manager

> List and toggle DeepSeek Harness filesystem skills — from the agent, or from a Settings page ("技能" / "Skills").

DeepSeek Harness loads skills straight from the filesystem (~/.agents/skills/<name>/SKILL.md, ~/.dsh/skills, project-level .dsh/skills and .agents/skills) but ships no management surface: there is no way to see which skills are installed, and no way to turn one off. dsh-skill-manager fills that gap.

  • skills_list — list every filesystem skill with its enabled state and one-line description.
  • skills_toggle — enable/disable any skill instantly (no restart).
  • Settings UI — a "技能" (Skills) page in Settings with the same list and enable/disable toggles (bundle-included, no extra setup).

Disabling renames <name>/SKILL.md to <name>/SKILL.md.disabled; enabling renames it back. The built-in skill-filesystem provider watches the roots, so a disabled skill disappears from the model-visible skill catalog immediately (and reappears when re-enabled). Skill content is never modified — only the SKILL.md filename.

Install

The plugin is a dsh bundle (declares dsh.bundle.patch), so it installs like any other profile plugin.

Requirements

  • Node.js ≥ 22.19 (check with node --version)
  • A dsh web profile (dsh plugin commands target the profile directory, e.g. --profile web)

⚠️ Name collision warning

There is an unrelated npm package also named dsh-skill-manager (published by a different author). Installing by bare npm name may pull the wrong package. Always install from this repository explicitly:

# from GitHub
dsh plugin --profile web add Thomas-key/dsh-skill-manager

# or from a local checkout
dsh plugin --profile web add ./dsh-skill-manager

Installing

Your profile directory is a pnpm workspace root, so the install command must include the -w flag (otherwise pnpm refuses with ERR_PNPM_ADDING_TO_ROOT):

dsh plugin --profile web add -w Thomas-key/dsh-skill-manager

For a local checkout, the plugin's runtime dependency (@deepseek-ai/dsh-tools) must be installed inside the checkout before adding it (the profile links the directory and Node resolves imports from it; a missing dependency crashes dsh web at startup with ERR_MODULE_NOT_FOUND):

cd dsh-skill-manager
npm install
cd ..
dsh plugin --profile web add -w ./dsh-skill-manager

Verify

Restart dsh web, then either:

  • open Settings → 技能 (the Settings UI lists every skill with enable/disable toggles), or
  • ask the agent: "List my skills." / "Disable the skill read."

The agent will call skills_list / skills_toggle for you. Toggling takes effect immediately — you can verify in the same session by asking what skills are visible. If dsh web fails to start after installing, check the error is not a missing @deepseek-ai/dsh-tools (see above).

Settings UI

The same management surface is available as a settings page ("技能" / "Skills" in Settings). The web UI ships as a client bundle (lib/client.js, loaded via the shell's __ModuleLoader__) and works out of the box — no extra setup, survives restarts.

> Legacy note: an earlier dynamic-plugin variant (plugin/skill-manager.js) provided the same UI via cordis_define/cordis_run. It is kept for reference only; the bundle now covers both the tools and the UI.

How it works

agent → skills_list / skills_toggle (model tools)
     → Host plugin → `node -e` bridge (stdin: JSON op, stdout: JSON result)
     → scans skill roots / renames SKILL.md <-> SKILL.md.disabled
     → dsh-skill-filesystem watcher → catalog updates instantly

The bridge is zero-dependency (Node built-ins only) and runs inside DSH's normal subprocess service.

Boundaries

  • Scans only the four standard skill roots; anything else is untouched.
  • Only renames SKILL.md/SKILL.md.disabled — never reads, edits, or deletes skill content.
  • Toggling a skill only hides it from the catalog (model-visible list). Installed files stay in place; disabling is fully reversible.

Maintenance status

This plugin was authored by deepseek-v4-flash (an AI agent). If this notice has not been removed, the author does not actively maintain this plugin — it was built to scratch a personal itch (DSH ships no skill-management surface) and works as-is. No updates are planned as long as it keeps working. Forks and PRs are welcome.

License

MIT