mxcli-olc-setup 1.0.0 → 1.0.1

package/README.md

@@ -1,6 +1,6 @@
 # mxcli-olc-setup
 
-Automates [mxcli](https://github.com/mendixlabs/mxcli) setup for Mendix projects. Downloads the correct mxcli binary for your platform, initializes it, adds AI-agent skills, and configures your project's `.gitignore`.
+Automates [mxcli](https://github.com/mendixlabs/mxcli) setup for Mendix projects. Downloads the correct mxcli binary for your platform, initializes it with all supported AI tools, adds AI-agent skills, and sets up a shared knowledge base that AI agents maintain across sessions.
 
 ## Quick Start
 
@@ -18,11 +18,34 @@ npx mxcli-olc-setup /path/to/mendix-project
 
 ## What It Does
 
-1. **Downloads mxcli** — fetches the latest release binary for your OS/architecture
-2. **Runs `mxcli init`** — initializes mxcli with Claude and OpenCode tool support
-3. **Adds Mendix Developer Skill** — copies the AI skill file to `.ai-context/skills/`
+1. **Downloads mxcli** — fetches the latest release binary for your OS/architecture into `.tools/mxcli/`
+2. **Runs `mxcli init --all-tools`** — initializes mxcli with all supported AI tools (Claude, OpenCode, Cursor, Windsurf, Continue.dev, Aider)
+3. **Adds Mendix Developer Skill** — copies the AI skill file to `.ai-context/skills/` with guardrails for `.mpr` inspection, microflow tracing, SCSS conventions, and more
 4. **Updates `.gitignore`** — appends entries for AI/mxcli generated files
-5. **Creates knowledge base** — generates a `project-knowledge-base.md` template for AI agents
+5. **Creates knowledge base** — generates a `project-knowledge-base.md` template where AI agents document what they learn about the project
+6. **Wires up knowledge base instructions** — appends instructions to `AGENTS.md` and `CLAUDE.md` telling AI agents to read and update the knowledge base every session
+
+## How the Knowledge Base Works
+
+After setup, a `project-knowledge-base.md` file lives at the project root. AI agents are instructed (via `AGENTS.md`, `CLAUDE.md`, and the Mendix Developer Skill) to:
+
+- **Read it** at the start of every session to learn what previous agents discovered
+- **Update it** when they find new information — module purposes, entity relationships, microflow traces, architectural patterns, gotchas
+
+This creates a persistent memory across AI sessions, so agents don't re-discover the same things.
+
+## CLI Usage
+
+```
+npx mxcli-olc-setup [project-path]
+
+Arguments:
+  project-path    Path to the Mendix project root (default: current directory)
+
+Options:
+  --help, -h      Show help
+  --version, -v   Show version
+```
 
 ## Global Install
 
@@ -44,3 +67,7 @@ mxcli-olc-setup
 | Windows | x64, arm64 |
 | macOS | x64 (Intel), arm64 (Apple Silicon) |
 | Linux | x64, arm64 |
+
+## Re-running
+
+Safe to run multiple times. It skips steps that are already done (existing binary, existing knowledge base) and only appends new `.gitignore` entries or knowledge base instructions if they're missing.

package/assets/mendix-developer-skill.md

@@ -172,6 +172,21 @@ Purpose: align decision dropdown spacing and sizing in the record sheet action r
 
 Existing legacy styles that already target `.mx-name-*` selectors should not be expanded unless the task is specifically to clean up or migrate them.
 
+## Project Knowledge Base
+
+A file named `project-knowledge-base.md` exists at the project root. This is a shared memory across AI sessions. You MUST update it whenever you learn something that would help a future AI agent (or yourself in a new session) work on this project.
+
+Update the knowledge base when you:
+
+1. Discover a module's purpose or responsibility.
+2. Identify key domain entities and their relationships.
+3. Trace an important microflow or page navigation path.
+4. Find a gotcha, edge case, or non-obvious constraint.
+5. Make an architectural decision or observe an existing pattern.
+6. Complete any analysis that produced reusable findings.
+
+When updating, add to the appropriate section (Module Map, Key Entities, Important Microflows, etc.). Add a dated entry under Change Log. Keep entries concise and factual. Do not remove existing entries unless they are confirmed wrong.
+
 ## Git ignore
 
 Whenever AI or mxcli generated any files in the project directory, make sure the files to be included in git ignore file of the Mendix project.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mxcli-olc-setup",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Automates mxcli setup for Mendix projects — downloads mxcli binary, runs init, adds custom skills, and configures gitignore",
   "bin": {
     "mxcli-olc-setup": "./setup.js"
@@ -13,6 +13,13 @@
   "engines": {
     "node": ">=14"
   },
-  "keywords": ["mendix", "mxcli", "ai", "low-code", "cli", "setup"],
+  "keywords": [
+    "mendix",
+    "mxcli",
+    "ai",
+    "low-code",
+    "cli",
+    "setup"
+  ],
   "license": "UNLICENSED"
 }

package/setup.js

@@ -68,10 +68,11 @@ function printHelp() {
 
   What it does:
     1. Downloads the latest mxcli binary for your platform
-    2. Runs mxcli init with Claude and OpenCode tools
+    2. Runs mxcli init with all supported AI tools
     3. Adds the Mendix Developer Skill to .ai-context/skills/
     4. Appends AI/mxcli entries to .gitignore
     5. Creates a project-knowledge-base.md template
+    6. Wires up knowledge base instructions in AGENTS.md and CLAUDE.md
 `);
 }
 
@@ -232,7 +233,7 @@ function runMxcliInit(binaryPath, projectRoot) {
   log('Running mxcli init...');
   try {
     const result = execSync(
-      `"${binaryPath}" init --tool claude --tool opencode "${projectRoot}"`,
+      `"${binaryPath}" init --all-tools "${projectRoot}"`,
       { encoding: 'utf8', stdio: 'pipe', cwd: projectRoot }
     );
     log('mxcli init completed successfully');
@@ -333,6 +334,34 @@ function createKnowledgeBase(projectRoot) {
   log('Created project-knowledge-base.md');
 }
 
+const KNOWLEDGE_BASE_NOTICE = `
+
+## Project Knowledge Base
+
+A shared knowledge base exists at \`project-knowledge-base.md\` in the project root. This file persists across AI sessions.
+
+**You MUST update it** when you discover new information about the project — module purposes, entity relationships, microflow traces, architectural patterns, gotchas, or any findings that would help a future agent. Add a dated entry under the Change Log section.
+
+Read this file at the start of every session to understand what previous agents have already learned.
+`;
+
+function appendKnowledgeBaseNotice(projectRoot) {
+  const targets = ['AGENTS.md', 'CLAUDE.md'];
+  for (const filename of targets) {
+    const filePath = path.join(projectRoot, filename);
+    if (!fs.existsSync(filePath)) continue;
+
+    const content = fs.readFileSync(filePath, 'utf8');
+    if (content.includes('project-knowledge-base.md')) {
+      log(`${filename} already references knowledge base, skipping`);
+      continue;
+    }
+
+    fs.appendFileSync(filePath, KNOWLEDGE_BASE_NOTICE, 'utf8');
+    log(`Added knowledge base instructions to ${filename}`);
+  }
+}
+
 async function main() {
   const projectRoot = getProjectRoot();
   log(`Project root: ${projectRoot}`);
@@ -358,6 +387,8 @@ async function main() {
 
   createKnowledgeBase(projectRoot);
 
+  appendKnowledgeBaseNotice(projectRoot);
+
   log('Setup complete!');
 }