Merge pull request #2 from namos2502/dev

Enhance setup instructions and improve auto guard functionality

Author: namos2502

.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "xflow",
   "description": "Cross-agent workflow plugin — gives your agent the full programmatic reference for every CLI tool you use. One plugin. Every CLI. Stay in flow.",
-  "version": "0.5.2",
+  "version": "0.5.3",
   "author": {
     "name": "namos2502"
   },

CHANGELOG.md

@@ -2,6 +2,12 @@
 
 All notable changes to xFlow will be documented here.
 
+## [0.5.3] — 2026-03-21
+
+### Fixed
+- `/xflow:auto` guard restructured from a soft "Step 1" bullet to a hard `⛔ STOP` block — models were reading past the conditional and activating regardless; now the gate is unambiguous and precedes the activation section entirely
+- `/xflow:setup` now ends every confirmation with a prompt to run `/xflow:auto` next — closes the gap where users had no clear signal of what to do after setup
+
 ## [0.5.2] — 2026-03-21
 
 ### Added

README.md

@@ -11,6 +11,8 @@ xFlow fixes that.
 
 **xFlow** is a cross-agent orchestration plugin for your AI agent CLI. It turns your active AI agent into a persistent **control center** that decomposes tasks, delegates to other CLI agents, and manages the full workflow — while you stay focused on what you're building.
 
+> **Note:** xFlow is designed for use with 2 or more AI CLI platforms. To get the most out of cross-agent workflows, you need access to at least two supported CLI agents (e.g. Copilot CLI + Claude Code).
+
 ## How it works
 
 xFlow uses a **Control Center → Agents** model:
@@ -36,8 +38,6 @@ No context bloat. No manual coordination. The workflow just happens.
 
 **Reduce your token spend.** Every delegated subtask runs in a separate subprocess. Only the final report comes back into your context — no reasoning chains, no intermediate tool calls.
 
-**Extend without friction.** Adding a new CLI agent = drop a `skills/agents/<name>/SKILL.md` file. No other changes needed.
-
 ## Get started
 
 Before installing, make sure you have at least one supported CLI agent set up.
@@ -140,4 +140,4 @@ The orchestration skill teaches your CLI when and how to delegate:
 
 ## License
 
-MIT
+This project is licensed under the MIT License

commands/auto.md

@@ -2,18 +2,21 @@
 description: "Activate autonomous multi-agent mode — Claude will know when and how to invoke copilot -p or claude -p for the rest of this session"
 ---
 
-**Step 1 — Check if setup has been run:**
+⛔ **BEFORE ANYTHING ELSE — verify setup is complete:**
 
 Read `~/.claude/CLAUDE.md` and `~/.copilot/copilot-instructions.md`. Check whether each contains a `## xFlow` section.
 
-- If **neither** file has the section — stop and tell the user:
-  > ⚠️ xFlow is not set up yet. Run `/xflow:setup` first to detect your CLI agents and register xFlow, then run `/xflow:auto` again.
+If **neither** file contains a `## xFlow` section:
+> **STOP. Do not load skills. Do not activate. Do not proceed.**
+> Tell the user: ⚠️ xFlow is not set up yet. Run `/xflow:setup` first to detect your CLI agents and register xFlow, then re-run `/xflow:auto`.
 
-- If **only one** file has the section — warn the user setup may be incomplete and suggest re-running `/xflow:setup`, then proceed to Step 2.
+If **only one** file has the section — warn the user that setup may be incomplete and suggest re-running `/xflow:setup`. Then proceed.
 
-- If **both** files have the section — proceed to Step 2.
+If **both** files have the section — proceed.
 
-**Step 2 — Load skills and activate:**
+---
+
+**Load skills and activate:**
 
 Read and load the following skill files into your context for this session:
 

commands/setup.md

@@ -110,3 +110,6 @@ Fill in the actual agent status from Step 1 before writing.
 
 - ✅ All installed agents authenticated and both instruction files updated — tell the user xFlow is ready and always-on in Claude Code and Copilot CLI.
 - ✅ Instruction files already had the section — tell the user xFlow was already configured. Offer to re-run auth check if needed.
+
+In all cases, end with:
+> 🚀 Run `/xflow:auto` to activate multi-agent mode for this session.