Add ai-tools-setup, computer-health-check, knowledge-base, obsidian-v…

plugins/claude/.claude-plugin/plugin.json

@@ -1,6 +1,7 @@
 {
   "name": "desktop-commander",
   "displayName": "Desktop Commander",
+  "version": "0.1.0",
   "description": "MCP server for terminal commands, process management, and file operations across text, code, PDF, DOCX, Excel, images, and structured data",
   "author": {
     "name": "Desktop Commander",

plugins/claude/README.md

@@ -1,14 +1,19 @@
 # Desktop Commander Claude Code Plugin
 
-Desktop Commander gives Claude Code access to a local MCP server for terminal sessions, filesystem work, structured documents, search, process management, and SSH workflows.
+Desktop Commander gives Claude Code access to a local MCP server for terminal sessions, filesystem work, structured documents, search, process management, and SSH workflows — plus a set of skills that steer the agent toward those capabilities for common tasks.
 
 ## Components
 
-- Skill: `desktop-commander-overview`, available as `/desktop-commander:desktop-commander-overview`, explains when to use Desktop Commander MCP tools and how common workflows compose.
 - MCP server: `desktop-commander`, installed with `npx -y @wonderwhy-er/desktop-commander@latest`.
+- Skill: `desktop-commander-overview` — explains when to use Desktop Commander MCP tools and how common workflows compose.
+- Skill: `terminal` — persistent shells/REPLs, SSH, Windows PowerShell, process/port inspection, and saving recurring workflows as scripts.
+- Skill: `computer-health-check` — read-only system health check (CPU, memory, disk, battery, startup) with safe, opt-in cleanups.
+- Skill: `ai-tools-setup` — install, configure, and repair Claude Desktop, MCP servers, and local AI tooling (mcp.json / claude_desktop_config.json).
+- Skill: `knowledge-base` — build and maintain a Markdown knowledge base any AI agent can navigate without a vector database.
+- Skill: `obsidian-vault` — organize Obsidian vaults: MOCs, wikilinks, properties, Dataview/Bases dashboards, and orphan/link cleanup.
 
 ## Usage
 
-Install the plugin in Claude Code, then ask Claude to use Desktop Commander when it needs persistent shells, long-running processes, local file reads, Excel/DOCX/PDF handling, large CSV analysis, or remote SSH sessions.
+Install the plugin in Claude Code, then ask Claude to use Desktop Commander when it needs persistent shells, long-running processes, local file reads, Excel/DOCX/PDF handling, large CSV analysis, remote SSH sessions, a system health check, AI-tooling setup, or knowledge-base/Obsidian organization.
 
 For full project documentation, see the repository README.

plugins/claude/skills/ai-tools-setup/SKILL.md

@@ -0,0 +1,107 @@
+---
+name: ai-tools-setup
+version: 0.1.0
+audience: agent
+description: >-
+  Set up, connect, validate, and repair Claude Desktop and MCP servers using
+  Desktop Commander — inspecting and fixing mcp.json / claude_desktop_config.json,
+  getting MCP servers to connect, and diagnosing local AI tooling. Use when a
+  user wants to install, configure, connect, validate, or fix Claude Desktop, an
+  MCP server, an MCP config file, or a local AI agent/gateway. Trigger on the
+  symptoms that actually bring people in: "Claude can't see my tools", "my MCP
+  server isn't showing up", "tools aren't loading in Claude", "add an MCP
+  server", "my MCP isn't connecting", "fix my Claude Desktop config", "the agent
+  won't respond", "wire up Ollama / LM Studio", "restart Claude Desktop". Also
+  covers local agent gateways like OpenClaw and Hermes.
+---
+
+# AI Tools Setup Assistant
+
+Help users install, configure, and repair AI tooling on their own machine using
+Desktop Commander's file and terminal tools. The work is usually
+**configure-and-troubleshoot**, not clean installs — assume something already
+exists and may be half-broken.
+
+## Golden rules (read before doing anything)
+
+1. **Detect the OS and shell first.** Read `get_config` and use its `systemInfo`
+   and `allowedDirectories` — file paths, commands, and config locations differ
+   across macOS, Windows, and Linux, so confirm the platform before assuming
+   anything.
+2. **Verify the model/API key works before you build.** The most common reason
+   these sessions dead-end is a provider auth or billing error discovered
+   *after* a lot of setup work. Make one tiny test call with `start_process` (a quick
+   curl to the provider) early, and tell the user plainly if the key is invalid
+   or out of credit.
+3. **Translate errors, don't dump them.** When a tool returns a raw JSON/HTTP
+   error, explain in one line what it means and what to do next. If a model is
+   missing, deprecated, rate-limited, or doesn't support tool-calling, suggest a
+   concrete working alternative instead of retrying the same thing.
+4. **Never echo secrets.** API keys, tokens, and passwords go *into the right
+   config or `.env` file* — never repeated back in chat. If a user pastes a live
+   secret, warn them, use it, and suggest they rotate it if it was shared
+   insecurely. Redact secrets in any output, logs, or summaries you produce.
+5. **Back up before you edit.** Copy the config to a `.bak` first with
+   `start_process` (`cp file.json file.json.bak`, or `copy` on Windows), then make
+   the change with `edit_block` so existing entries survive — reach for
+   `write_file` only on a from-scratch config. Validate the JSON after every edit
+   by parsing it with `start_process`.
+6. **Verify, then stop.** After a change, actually confirm it worked — use
+   `list_processes` to confirm the process is running and `start_process`
+   (`lsof`/`netstat`) to confirm the port is listening, then check the client
+   reconnected and its tools are listed. Don't declare success on a guess.
+7. **Make it repeatable.** Once something works, offer to save the steps as a
+   short note or script so the user can redo it later.
+
+## What this skill covers
+
+This skill routes to one of three reference files depending on the task. Read
+the relevant file before acting — they contain the exact paths, commands, and
+known failure modes.
+
+- **Claude Desktop & MCP servers** — installing Claude Desktop, adding MCP
+  servers, and inspecting / validating / repairing MCP config files
+  (`claude_desktop_config.json`, `.claude/mcp.json`, `~/.lmstudio/mcp.json`,
+  and similar). Also connection health checks.
+  → Read `references/claude-desktop-mcp.md`
+
+- **OpenClaw** — a self-hosted multi-channel gateway for AI agents. Install,
+  onboard, wire a model/provider, connect channels (Telegram, Slack, iMessage,
+  etc.), diagnose the gateway, and uninstall.
+  → Read `references/openclaw.md`
+
+- **Hermes** — disambiguate first: **Hermes Agent** (the Nous Research
+  self-improving agent framework) vs. **Hermes 3** (just a local LLM you run via
+  Ollama). Install, run `hermes setup --portal`, configure providers, personality
+  (`SOUL.md`), memory/skills, MCP, and messaging.
+  → Read `references/hermes.md`
+
+When in doubt about a tool's current behavior, fetch the official docs:
+OpenClaw — https://docs.openclaw.ai/ ·
+Hermes — https://hermes-agent.nousresearch.com/docs/
+
+## General workflow
+
+1. **Clarify the goal and platform.** Install vs. configure vs. fix? Which OS?
+   Which tool/client?
+2. **Inspect current state.** `read_file` the relevant config files and use
+   `list_processes` to check whether the relevant process/port/service is already
+   running before changing anything.
+3. **Read the matching reference file** for exact paths and commands.
+4. **Make the smallest change that moves forward**, backing up configs first.
+5. **Verify** the result concretely.
+6. **Summarize** what changed, where, and how to undo it — with secrets redacted.
+
+## A generic "my agent isn't responding" diagnostic
+
+Use this order for OpenClaw, Hermes, or any local agent/gateway:
+
+1. **Provider/auth** — is a valid API key configured for the active provider,
+   and does it have credit? (Most failures stop here.)
+2. **Model** — does the chosen model exist, and does it support tool-calling?
+   Local models especially often don't.
+3. **Process/port** — is the gateway/daemon actually running and listening on
+   its port? Is another process holding that port?
+4. **Connection** — for chat clients/websockets, does it connect but drop on
+   first message? That usually points back to provider/model, not networking.
+5. **Logs** — read the tool's logs for the first real error, not the last one.

plugins/claude/skills/ai-tools-setup/references/claude-desktop-mcp.md

@@ -0,0 +1,98 @@
+# Claude Desktop & MCP Servers
+
+How to install Claude Desktop, add MCP servers, and inspect / validate / repair
+MCP configuration files. This is the most common and most mechanical task — high
+success rate when done carefully.
+
+## Config file locations
+
+MCP-capable clients each keep their own config. Find the right one first.
+
+| Client | macOS | Windows |
+|---|---|---|
+| Claude Desktop | `~/Library/Application Support/Claude/claude_desktop_config.json` | `%APPDATA%\Claude\claude_desktop_config.json` |
+| Claude Code (per project) | `<project>/.claude/mcp.json` or `.mcp.json` | same |
+| LM Studio | `~/.lmstudio/mcp.json` | `%USERPROFILE%\.lmstudio\mcp.json` |
+| Cursor | `~/.cursor/mcp.json` | `%USERPROFILE%\.cursor\mcp.json` |
+
+If the file doesn't exist yet, create it with a minimal valid skeleton:
+
+```json
+{ "mcpServers": {} }
+```
+
+## Anatomy of an MCP server entry
+
+A stdio server entry has a launch `command`, `args`, and optional `env`:
+
+```json
+{
+  "mcpServers": {
+    "example": {
+      "command": "npx",
+      "args": ["-y", "@scope/some-mcp-server"],
+      "env": { "SOME_API_KEY": "..." }
+    }
+  }
+}
+```
+
+Notes:
+- On Windows, `npx` sometimes must be invoked as `cmd /c npx ...` — if a server
+  fails to launch on Windows but works on macOS, try wrapping the command.
+- Use absolute paths for local scripts; `~` is not expanded inside JSON.
+- Secrets belong in `env` here (or the server's own config), never pasted into
+  chat.
+
+## Add a server end-to-end
+
+1. **Confirm prerequisites.** Most MCP servers need Node (check `node -v`; many
+   need 18+). Some need Python or Docker. Install the runtime if missing.
+2. **Back up the config** (`claude_desktop_config.json` →
+   `claude_desktop_config.json.bak`).
+3. **Add the server block** under `mcpServers`. Preserve existing entries.
+4. **Validate the JSON** (parse it; check for trailing commas, unescaped
+   backslashes in Windows paths, missing quotes).
+5. **Fully restart the client.** Claude Desktop must be quit and reopened — a
+   window reload is not enough.
+6. **Verify.** Confirm the server appears and its tools are listed. If the
+   client shows a "failed" server, read its MCP log.
+
+## Inspect, validate, repair (the core job)
+
+When a user says MCP "isn't working," diagnose in this order:
+
+1. **Read the config file** and pretty-print it. Most breakages are here.
+2. **Validate JSON.** The frequent culprits:
+   - trailing comma after the last entry
+   - single backslashes in Windows paths (need `\\` or forward slashes)
+   - smart quotes from copy-paste instead of straight quotes
+   - a server block pasted at the wrong nesting level
+3. **Check the launch command resolves** — does `command` exist on PATH? Run it
+   manually in a terminal to see the real error.
+4. **Check credentials** — is the required `env` key present and valid?
+5. **Repair, back up first, restart the client, then re-verify.**
+
+## Connection health check
+
+Separate the three failure classes so you fix the right thing:
+
+- **Config typo** → client shows the server as failed immediately on start; the
+  JSON or command is wrong. Fix the file.
+- **Server down / crashes** → server starts then exits; running the command
+  manually reproduces the crash. Fix deps/runtime.
+- **Auth failure** → server starts and lists tools, but calls fail with 401/403.
+  Fix the API key/token.
+
+## Claude Desktop install
+
+If Claude Desktop itself isn't installed, point the user to the official
+download and confirm the app launches before touching MCP config. Don't
+side-load from unofficial mirrors.
+
+## Safety
+
+- Only add MCP servers from sources the user trusts; an MCP server runs with the
+  user's privileges.
+- Back up before editing; redact secrets in summaries; restart the client to
+  apply changes.

plugins/claude/skills/ai-tools-setup/references/hermes.md

@@ -0,0 +1,90 @@
+# Hermes
+
+**Disambiguate first — "Hermes" means two different things:**
+
+1. **Hermes Agent** — the self-improving AI agent framework built by Nous
+   Research (a learning loop that creates/improves skills, persistent memory,
+   runs across many platforms). This is almost always what users mean when they
+   say "set up Hermes." Docs: https://hermes-agent.nousresearch.com/docs/
+2. **Hermes 3 (and other Hermes models)** — Nous Research's open-weight LLMs.
+   "Set up Hermes 3" usually just means running a local model: `ollama pull
+   hermes3` then `ollama run hermes3`. If that's the goal, treat it as ordinary
+   local-model setup and stop here.
+
+The rest of this file covers **Hermes Agent**. MIT licensed. For current
+commands, fetch the docs (there's also `/docs/llms.txt` for a compact index).
+
+## Install
+
+**Desktop app (Windows/macOS):** download the Hermes Desktop installer from
+https://hermes-agent.nousresearch.com/ and run it (installs CLI + desktop app).
+
+**CLI only:**
+
+```bash
+# Linux / macOS / WSL2 / Android (Termux)
+curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
+```
+
+```powershell
+# Windows (native), in PowerShell
+iex (irm https://hermes-agent.nousresearch.com/install.ps1)
+```
+
+See the full Installation Guide for what the installer does, per-user vs. root
+layout, and Windows notes.
+
+## Fastest path to a working agent
+
+```bash
+hermes setup --portal
+```
+
+One OAuth via **Nous Portal** covers a model plus the four Tool Gateway tools
+(web search, image generation, TTS, browser). This is the smoothest first run —
+recommend it unless the user specifically wants their own provider keys.
+
+## Providers & config
+
+Hermes works with **Nous Portal, OpenRouter, OpenAI, or any compatible
+endpoint**. Configuration lives in Hermes' config file (see the Configuration
+docs). Common requests:
+
+- Wire OpenRouter / OpenAI / a custom endpoint with the user's own key — verify
+  the key works before going further.
+- "Use my workspace Gemini / no API cost" and similar — set the matching
+  provider/model in config; confirm the model ID is current and supported.
+- If a Portal login code is rejected, re-run `hermes setup --portal` and check
+  the user is pasting the current code.
+
+## Key features to configure
+
+- **Personality (`SOUL.md`)** — defines the agent's default voice/tone. Users
+  often ask to generate or update it. Keep it a plain Markdown persona file.
+- **Memory** — persistent, grows across sessions; mostly automatic.
+- **Skills** — procedural memory the agent creates and reuses; compatible with
+  the open agentskills.io standard, so skills are portable.
+- **MCP integration** — Hermes can connect to MCP servers and filter their
+  tools. For config-file mechanics, see `claude-desktop-mcp.md`; for Hermes-side
+  specifics see the docs' MCP page.
+- **Messaging gateway** — Telegram, Discord, Slack, WhatsApp, Signal, Teams, and
+  20+ platforms from one gateway. Telegram is the quickest to validate.
+- **Backends** — Hermes can run on 6 terminal backends: local, Docker, SSH,
+  Daytona, Singularity, Modal. Daytona/Modal give serverless persistence (idle
+  ≈ free). Pick local for a first run; Docker if the user wants isolation.
+
+## Troubleshooting checklist
+
+1. **Install/update stuck** → check the installer/log output; confirm the
+   runtime and PATH; on Windows watch for permission/antivirus interference.
+   Re-run the installer if needed.
+2. **"Agent won't respond / got stuck"** → run the generic diagnostic: provider
+   key valid + has credit → model exists and supports tools → process running →
+   read logs for the first real error.
+3. **Portal/auth errors** → re-run `hermes setup --portal`; verify the account
+   and code.
+4. **Provider model errors (404/402/403)** → switch to a current, available
+   model.
+
+See the FAQ & Troubleshooting page in the docs for tool- and platform-specific
+fixes.