Add development environment setup (Gemfile, .gitignore, AGENTS.md)

.gitignore

@@ -0,0 +1,10 @@
+# Jekyll
+_site/
+.sass-cache/
+.jekyll-cache/
+.jekyll-metadata
+
+# Bundle
+vendor/
+Gemfile.lock
+.bundle/

AGENTS.md

@@ -0,0 +1,25 @@
+# AGENTS.md
+
+## Cursor Cloud specific instructions
+
+### Project Overview
+This is a bilingual (Korean/English) Jekyll static blog about data engineering. It has no CI/CD pipeline, no linter config, and no automated tests. The primary development loop is: edit Markdown/Liquid → build → preview.
+
+### Prerequisites
+- Ruby 3.x with `bundler` and `jekyll` gems (installed system-wide via `sudo gem install bundler jekyll`)
+- The `Gemfile` in the repo root declares project-level gem dependencies
+
+### Development Commands
+| Task | Command |
+|------|---------|
+| Install deps | `bundle install` (uses local `vendor/bundle` path via `.bundle/config`) |
+| Build site | `bundle exec jekyll build` |
+| Dev server | `bundle exec jekyll serve --host 0.0.0.0 --port 4000` |
+| Dev server (detached) | `bundle exec jekyll serve --host 0.0.0.0 --port 4000 --detach` |
+
+### Caveats
+- **No linter or test suite**: This project has no ESLint, RuboCop, or automated test framework. Validation is limited to `jekyll build` succeeding without errors.
+- **Liquid warnings are benign**: Build produces Liquid syntax warnings from dbt/Jinja template code embedded in blog post Markdown. These are cosmetic and do not affect the output.
+- **`_site/` is committed**: The pre-built output is checked in. After `jekyll build`, the `_site/` directory is regenerated.
+- **`vendor/bundle` is local**: Bundler is configured to install gems into `./vendor/bundle` (gitignored via `_config.yml` exclude). This avoids permission issues.
+- **No Gemfile.lock committed**: The lock file is generated locally by `bundle install`.

Gemfile

@@ -0,0 +1,6 @@
+source "https://rubygems.org"
+
+gem "jekyll", "~> 4.4"
+gem "kramdown", "~> 2.4"
+gem "rouge", "~> 4.0"
+gem "webrick", "~> 1.8"