claudemesh-cli 1.4.0 → 1.5.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claudemesh-cli",
-  "version": "1.4.0",
+  "version": "1.5.0",
   "description": "Peer mesh for Claude Code sessions — CLI + MCP server.",
   "keywords": [
     "claude-code",

package/skills/claudemesh/SKILL.md

@@ -24,6 +24,13 @@ If the parent Claude session was launched via `claudemesh launch`, an MCP push-p
 | `--mesh <slug>` | Target a specific mesh. Required when the user has multiple meshes joined. Default: first/only joined mesh, or interactive picker. |
 | `--json` | Emit JSON instead of human-readable text. Use this when you need to parse the output. |
 | `--json field1,field2` | Project specific fields (modeled on `gh --json`). Friendly aliases like `name` → `displayName` are resolved automatically. |
+| `--approval-mode <mode>` | `plan` / `read-only` deny all writes; `write` (default) prompts on destructive verbs from the policy file; `yolo` bypasses every prompt. |
+| `--policy <path>` | Override the policy file (default `~/.claudemesh/policy.yaml`, auto-created on first run). |
+| `-y` / `--yes` | Auto-approve any policy prompt. Equivalent to `--approval-mode yolo` for the current invocation. |
+
+## Policy & confirmation
+
+Every broker-touching verb runs through a policy gate before dispatch. The default policy allows reads and prompts on destructive writes (`peer kick/ban/disconnect`, `file delete`, `vector/vault delete`, `memory forget`, `skill remove`, `webhook delete`, `watch remove`, `sql/graph execute`, `mesh delete`). When you call `claudemesh` from a non-interactive context (cron, scripts, Claude's Bash tool), prompts auto-deny — pass `-y` or `--approval-mode yolo` for verbs you've vetted, or edit `~/.claudemesh/policy.yaml` to mark them `decision: allow`. Every gate decision is appended to `~/.claudemesh/audit.log` (newline-JSON).
 
 ## Resources and verbs