@jhm1909/ag-kit 0.1.0 → 0.2.1

package/.agent/rules/pre-commit-checklist.md

@@ -0,0 +1,40 @@
+---
+trigger: model_decision
+description: Always apply when doing or writing testing
+---
+
+# Pre-Commit Checklist
+
+> IMPORTANT: Run this checklist BEFORE every commit. Do not skip.
+
+## Verification Steps
+
+1. **Skill check**: Did you load the correct skill for this task?
+   - If unsure, re-read `.agent/rules/skill-loading.md`
+
+2. **Workflow check**: Did you follow the workflow completely?
+   - If no workflow was used, state why explicitly
+
+3. **Code quality**:
+   - [ ] No hardcoded secrets or API keys
+   - [ ] No `console.log` debugging left behind
+   - [ ] No commented-out code blocks
+   - [ ] No TODO without a linked issue
+
+4. **Testing** (if code changed):
+   - [ ] Tests pass (`npm test` or equivalent)
+   - [ ] New code has test coverage
+
+5. **Documentation** (if behavior changed):
+   - [ ] README or docs updated if user-facing behavior changed
+   - [ ] Code comments for non-obvious logic
+
+6. **Git hygiene**:
+   - [ ] `.env` files are NOT staged
+   - [ ] `node_modules/` is NOT staged
+   - [ ] Commit message follows Conventional Commits format
+
+## When to Skip
+
+- Typo-only fixes (docs, comments)
+- Dependency version bumps with no code changes

package/.agent/rules/workflow-advisor.md

@@ -5,13 +5,14 @@ description: Always suggest relevant workflows when user starts a new task or se
 
 # Workflow Advisor Rule
 
-## Critical Rules (MUST Follow)
+## Critical Rules (MANDATORY)
 
-1. **MUST** suggest relevant workflows when user describes a new project or task
+1. **MUST** suggest relevant workflows when user describes a new project or task. This is not optional.
 2. **MUST** suggest the *next* workflow when a current workflow completes
 3. **MUST** present workflow chains for multi-phase projects, not individual commands
-4. **MUST NOT** overwhelm — suggest max 3 options at a time
-5. **MUST NOT** force workflows — suggestions are optional, user can always say "no"
+4. **MUST** check `.agent/rules/skill-loading.md` before loading any skill
+5. **MUST NOT** overwhelm — suggest max 3 options at a time
+6. **MUST NOT** skip workflow steps to be "faster" — follow the full workflow
 
 ## When to Activate
 

package/.agent/templates/AGENTS.md

@@ -0,0 +1,45 @@
+This project uses [Antigravity Kit](https://github.com/jhm1909/antigravity-kit) for AI agent configuration.
+
+## Mandatory Rules
+
+> IMPORTANT: You MUST follow these rules for EVERY task. No exceptions.
+
+1. BEFORE any task: read `.agent/rules/skill-loading.md` to load the correct skill
+2. BEFORE any commit: follow `.agent/workflows/commit.md` completely
+3. BEFORE writing code: check `.agent/rules/clean-code.md` for standards
+4. BEFORE writing tests: check `.agent/rules/testing.md` for requirements
+5. BEFORE writing docs: check `.agent/rules/documents.md` for conventions
+6. When starting a new task: suggest a workflow from `.agent/workflows/`
+
+## Structure
+
+All agent capabilities are in `.agent/`:
+
+- `skills/` — 21 domain expert skills (frontend, backend, AI, security, design, etc.)
+- `workflows/` — 15 chainable task workflows (brainstorm, implement, test, commit, etc.)
+- `rules/` — 10 always-on coding standards and guardrails
+- `skills-manifest.json` — Skill routing, triggers, and profiles
+
+## How to Use
+
+1. Read `.agent/skills-manifest.json` to understand available skills and their triggers
+2. When starting a new task, check `.agent/workflows/guide.md` for workflow discovery
+3. Skills are loaded on-demand — only read `SKILL.md` when the task matches a skill's triggers
+4. Follow rules in `.agent/rules/` at all times
+
+## Workflow Chains
+
+```
+New Project:  /brainstorm -> /documentation -> /break-tasks -> /implement-feature -> /commit
+New Feature:  /break-tasks -> /implement-feature -> /gen-tests -> /commit
+Bug Fix:      /development -> /gen-tests -> /commit
+Debug:        /debug -> /commit
+```
+
+## Self-Verification
+
+Before completing ANY task, verify:
+- [ ] Loaded the correct skill for this task type
+- [ ] Followed the workflow steps in order (did not skip steps)
+- [ ] Applied relevant rules (testing, docs, git, clean-code)
+- [ ] Code passes lint and tests

package/README.md

@@ -1,5 +1,5 @@
 <p align="center">
-  <img src="https://img.shields.io/badge/Antigravity_Kit-v3.0-8B5CF6?style=for-the-badge&labelColor=1a1a2e" alt="Antigravity Kit" />
+  <img src="https://img.shields.io/badge/Antigravity_Kit-v0.2.0-8B5CF6?style=for-the-badge&labelColor=1a1a2e" alt="Antigravity Kit" />
 </p>
 
 <h1 align="center">Antigravity Kit</h1>
@@ -43,11 +43,15 @@ Most AI coding assistants have **zero memory** between sessions. They forget you
 ## Quick Start
 
 ```bash
-# Install everything
+# npm
 npx @jhm1909/ag-kit init
 
-# Or pick a profile
+# pnpm
+pnpm dlx @jhm1909/ag-kit init
+
+# With a profile
 npx @jhm1909/ag-kit init --profile fullstack-saas
+pnpm dlx @jhm1909/ag-kit init --profile fullstack-saas
 ```
 
 That's it. Your `.agent/` directory is now populated with skills, workflows, and rules.
@@ -153,6 +157,8 @@ Install only what you need:
 
 ```bash
 npx @jhm1909/ag-kit init --profile <name>
+# or
+pnpm dlx @jhm1909/ag-kit init --profile <name>
 ```
 
 | Profile | Included Skills | Optional |

package/cli/index.js

@@ -14,7 +14,7 @@ const fs = require('fs');
 const path = require('path');
 
 // ─── Constants ───────────────────────────────────────────────────────
-const VERSION = '0.1.0';
+const VERSION = '0.2.1';
 const KIT_NAME = '@jhm1909/ag-kit';
 const AGENT_DIR = '.agent';
 
@@ -67,8 +67,8 @@ function copyDir(src, dest, allowedDirs = null, depth = 0) {
     const srcPath = path.join(src, entry.name);
     const destPath = path.join(dest, entry.name);
 
-    // Skip tmp/ directory always
-    if (entry.name === 'tmp') continue;
+    // Skip tmp/ and templates/ directories always
+    if (entry.name === 'tmp' || entry.name === 'templates') continue;
 
     // At skills level (depth 1, parent is "skills"), filter by allowed dirs
     if (depth === 1 && allowedDirs && entry.isDirectory()) {
@@ -181,6 +181,14 @@ function cmdInit(args) {
   log(`  ${c.cyan}|${c.reset} Workflows   ${c.dim}Chainable task automations${c.reset}`);
   log(`  ${c.cyan}|${c.reset} Rules       ${c.dim}Coding standards & guardrails${c.reset}`);
   log(`  ${c.cyan}|${c.reset} Manifest    ${c.dim}Skill routing & profiles${c.reset}`);
+  // Copy AGENTS.md to project root for cross-platform discovery
+  const agentsMdSrc = path.join(SOURCE_AGENT_DIR, 'templates', 'AGENTS.md');
+  const agentsMdDest = path.join(targetDir, 'AGENTS.md');
+  if (fs.existsSync(agentsMdSrc) && (!fs.existsSync(agentsMdDest) || force)) {
+    fs.copyFileSync(agentsMdSrc, agentsMdDest);
+    success('Created AGENTS.md (cross-platform agent config)');
+  }
+
   log();
   log(`${c.dim}  Works with: Claude Code, Cursor, Gemini CLI, GitHub Copilot${c.reset}`);
   log();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jhm1909/ag-kit",
-  "version": "0.1.0",
+  "version": "0.2.1",
   "description": "AI Agent Kit — 21 skills, 15 workflows, manifest-driven routing for AI-powered coding",
   "main": "cli/index.js",
   "bin": {
@@ -15,6 +15,7 @@
     ".agent/ARCHITECTURE.md",
     ".agent/known-registries.json",
     ".agent/mcp_config.json",
+    ".agent/templates/",
     "README.md",
     "LICENSE"
   ],