cursor-guard 2.1.1 → 4.0.0

package/README.md

@@ -20,6 +20,10 @@ When Cursor's AI agent edits your files, there's a risk of accidental overwrites
 - **Configurable scope** — Protect only what matters via `.cursor-guard.json`
 - **Secrets filtering** — Sensitive files (`.env`, keys, certificates) are auto-excluded from backups
 - **Auto-backup script** — A cross-platform watcher (Node.js) that periodically snapshots to a dedicated Git branch without disturbing your working tree
+- **MCP tool calls (optional)** — 9 structured tools (diagnostics, snapshot, restore, status, dashboard, alerts, etc.) with JSON responses and lower token cost
+- **Auto-fix diagnostics** — `doctor_fix` automatically patches missing configs, uninitialized Git repos, gitignore gaps, and stale locks
+- **Proactive change-velocity alerts (V4)** — Auto-detects abnormal file change patterns and raises risk warnings
+- **Backup health dashboard (V4)** — One-call comprehensive view: strategy, counts, disk usage, protection scope, health status
 
 ---
 
@@ -84,19 +88,32 @@ After installation, your directory structure should look like this:
 
 ```
 .cursor/skills/cursor-guard/
-├── SKILL.md                            # AI agent instructions
+├── SKILL.md                            # AI agent instructions (with MCP dual-path logic)
+├── ROADMAP.md                          # Version evolution roadmap
 ├── README.md
 ├── README.zh-CN.md
 ├── LICENSE
 ├── package.json
 └── references/
     ├── lib/
-    │   ├── auto-backup.js              # Backup core (Node.js)
-    │   ├── guard-doctor.js             # Health check core
-    │   └── utils.js                    # Shared utilities
+    │   ├── auto-backup.js              # Backup watcher (calls Core)
+    │   ├── guard-doctor.js             # Health check CLI (calls Core)
+    │   ├── utils.js                    # Shared utilities
+    │   └── core/                       # V3 Core layer (pure logic)
+    │       ├── doctor.js               # Diagnostics (incl. MCP self-check)
+    │       ├── doctor-fix.js           # Auto-fix common issues
+    │       ├── snapshot.js             # Git snapshots + shadow copies
+    │       ├── backups.js              # Backup listing + retention
+    │       ├── restore.js              # Single file / project restore
+    │       ├── status.js               # Backup system status
+    │       ├── anomaly.js             # V4: Change-velocity detection
+    │       └── dashboard.js           # V4: Health dashboard aggregation
+    ├── mcp/
+    │   └── server.js                   # MCP Server (9 tools)
     ├── bin/
-    │   ├── cursor-guard-backup.js      # CLI entry: npx cursor-guard-backup
-    │   └── cursor-guard-doctor.js      # CLI entry: npx cursor-guard-doctor
+    │   ├── cursor-guard-backup.js      # CLI: npx cursor-guard-backup
+    │   ├── cursor-guard-doctor.js      # CLI: npx cursor-guard-doctor
+    │   └── cursor-guard-mcp (server.js)# CLI: npx cursor-guard-mcp
     ├── auto-backup.ps1 / .sh           # Thin wrappers
     ├── guard-doctor.ps1 / .sh
     ├── recovery.md                     # Recovery commands
@@ -128,7 +145,22 @@ The skill activates automatically when the AI agent detects risky operations or
 cp .cursor/skills/cursor-guard/references/cursor-guard.example.json .cursor-guard.json
 ```
 
-5. **(Optional) Run auto-backup** in a separate terminal:
+5. **(Optional) Enable MCP tool calls** — add to `.cursor/mcp.json`:
+
+```jsonc
+{
+  "mcpServers": {
+    "cursor-guard": {
+      "command": "node",
+      "args": ["<skill-path>/references/mcp/server.js"]
+    }
+  }
+}
+```
+
+This gives the AI agent 9 structured tools (diagnostics, snapshot, restore, dashboard, alerts, etc.) with JSON responses — faster, more reliable, and lower token cost. Everything works without MCP too.
+
+6. **(Optional) Run auto-backup** in a separate terminal:
 
 ```bash
 npx cursor-guard-backup --path /my/project
@@ -167,8 +199,18 @@ Regardless of config, you can always override per-request:
 
 Run in a separate terminal while working in Cursor. Cross-platform — requires Node.js >= 18.
 
+Important:
+
+- You can run the command from any directory
+- But `--path` must point to the project root you want to protect
+- If you are already in the project root, `--path .` is fine
+- If you are not in the project root, do not use `--path .`; use the full target path instead
+
 ```bash
-# Via npx (after npm install)
+# If you are already in the project root
+npx cursor-guard-backup --path .
+
+# If you are not in the project root
 npx cursor-guard-backup --path /my/project
 npx cursor-guard-backup --path /my/project --interval 30
 
@@ -179,6 +221,14 @@ npx cursor-guard-backup --path /my/project --interval 30
 ./references/auto-backup.sh /my/project
 ```
 
+Wrong example:
+
+```bash
+# You are in some other directory
+# In that case --path . protects the current directory, not your real project
+npx cursor-guard-backup --path .
+```
+
 The script uses Git plumbing commands to snapshot to `refs/guard/auto-backup` — it never switches branches or touches your working index. The ref lives outside `refs/heads/` so `git push --all` won't push it. Supports `shadow` mode for non-Git directories.
 
 ### Health Check
@@ -247,6 +297,11 @@ The skill activates on these signals:
 - Time-based recovery: "restore to N minutes ago", "go back to yesterday"
 - Version-based recovery: "previous version", "go back N versions"
 - History issues: checkpoints missing, Timeline not working, save failures
+- Health check: "guard doctor", "check guard setup", "is MCP working"
+- Auto-fix: "guard fix", "fix config"
+- Backup status: "guard status", "is the watcher running", "last backup time"
+- Dashboard: "dashboard", "health overview", "backup summary"
+- Alerts: "any alerts?", "change velocity warning", "risk status"
 
 ---
 
@@ -254,9 +309,12 @@ The skill activates on these signals:
 
 | File | Purpose |
 |------|---------|
-| `SKILL.md` | Main skill instructions for the AI agent |
-| `references/lib/auto-backup.js` | Auto-backup core logic (Node.js) |
-| `references/lib/guard-doctor.js` | Health check core logic (Node.js) |
+| `SKILL.md` | Main skill instructions for the AI agent (with MCP dual-path) |
+| `ROADMAP.md` | Version evolution roadmap (V2-V7) |
+| `references/lib/core/` | Core layer: 8 pure-logic modules (doctor / doctor-fix / snapshot / backups / restore / status / anomaly / dashboard) |
+| `references/mcp/server.js` | MCP Server: 9 structured tools (optional) |
+| `references/lib/auto-backup.js` | Auto-backup watcher (calls Core) |
+| `references/lib/guard-doctor.js` | Health check CLI shell (calls Core) |
 | `references/lib/utils.js` | Shared utilities (config, glob, git, manifest) |
 | `references/bin/cursor-guard-backup.js` | CLI entry: `npx cursor-guard-backup` |
 | `references/bin/cursor-guard-doctor.js` | CLI entry: `npx cursor-guard-doctor` |