This repository owns the shared @jongminchung packages used by downstream projects.

• @jongminchung/ui: published UI primitives, a default neutral theme, shared Tailwind styles, and semantic tokens.

Public packages are published to GitHub Packages. Consumers need the @jongminchung scope mapped to https://npm.pkg.github.com and a classic PAT with read:packages, including for public package downloads. Keep the token in the environment rather than committing it to .npmrc.

@jongminchung:registry=https://npm.pkg.github.com
//npm.pkg.github.com/:_authToken=${GITHUB_PACKAGES_TOKEN}

The UI package provides safe default theme values. Apps may override semantic tokens and continue to own product components and behavior. See DESIGN_SYSTEM.md for the token contract, component ownership rules, and runtime exception policy.

• Repository documentation
• Maintenance guide
• Contributing guide

Repository-wide scripts are owned by the workspace root and are run from that directory:

bun run fmt
bun run check
bun run links:check
bun run audit

Always include run for package scripts so commands are visibly distinguished from Bun subcommands.

audit queries the registry advisory database and fails on high-severity dependency findings. It requires network access and is run manually during maintenance and before releases instead of being part of the offline-reproducible check chain.

links:check uses Podman when available, otherwise Docker, to run the pinned Lychee container with the repository mounted read-only. It checks local links and anchors in Markdown and HTML without making network requests.

Each workspace owns its build, typecheck, and test commands. Select one with a filter instead of adding a package-specific wrapper to the root manifest:

bun run --filter @jongminchung/web build

The manually triggered package workflow replaces @jongminchung/ui at 1.0.0. This is a mutable snapshot channel: the same version can have different API, contents, and integrity, so SemVer compatibility and lockfile reproducibility are not guaranteed. Consumers must force a new resolution, such as bun update --force <package>@1.0.0, and commit the resulting lockfile whenever they adopt a replacement.

The workflow installs the shared lockfile from the repository root, then lints, typechecks, and runs coverage and Node consumer checks for UI on Node 24 and 26. It builds the archive before deleting the fixed version, publishes that archive, and verifies registry integrity and consumer imports. See the release and recovery runbook. GitHub authentication is supplied only through the GH_PAT Actions secret.

bun install --frozen-lockfile --ignore-scripts
bun run --filter @jongminchung/ui typecheck
bun run --filter @jongminchung/ui test:coverage
bun run --filter @jongminchung/ui test:node
bun run --filter @jongminchung/ui publish:dry-run

I'm a Night 🦉

🌞 Morning                985 commits         ███░░░░░░░░░░░░░░░░░░░░░░   11.10 % 
🌆 Daytime                2135 commits        ██████░░░░░░░░░░░░░░░░░░░   24.05 % 
🌃 Evening                3366 commits        █████████░░░░░░░░░░░░░░░░   37.92 % 
🌙 Night                  2390 commits        ███████░░░░░░░░░░░░░░░░░░   26.93 % 

📊 This Week I Spent My Time On

💬 Programming Languages: 
Java                     6 hrs 40 mins       █████████░░░░░░░░░░░░░░░░   37.54 % 
YAML                     4 hrs 52 mins       ███████░░░░░░░░░░░░░░░░░░   27.41 % 
Other                    2 hrs 28 mins       ███░░░░░░░░░░░░░░░░░░░░░░   13.94 % 
Bash                     1 hr 10 mins        ██░░░░░░░░░░░░░░░░░░░░░░░   06.59 % 
Markdown                 49 mins             █░░░░░░░░░░░░░░░░░░░░░░░░   04.67 % 

🤖 AI Coding This Week

⏱ AI Coding Time: 9 hrs 23 mins (52.89%)

✍️ 399 lines written by AI, 1,202 lines written by hand (24.92% AI-written)

🔤 4,044,523 Input Tokens, 329,610 Output Tokens

💵 $36.27 Estimated AI Cost This Week

🧠 46 AI Sessions, 172 AI Prompts

GPT                      478 lines           █████████████████████████   100.00 % 
Codex-Vscode             0 lines             ░░░░░░░░░░░░░░░░░░░░░░░░░   00.00 % 
Opencode-Cli             0 lines             ░░░░░░░░░░░░░░░░░░░░░░░░░   00.00 % 

🔎 AI Coding Insights:
🧑‍💻 Mostly Hands-On — 24.92% of written lines came from AI
📄 Detailed Prompter — average 1,030 characters per prompt
🔁 Iterative Prompter — average 4 prompts per session
🔍 Hands-On Reviewer — 79.84% of changed lines were hand-edited

• 설치와 공통 make fmt·make lint 사용법은 기여 가이드를 참고함