# NomaDamas/k-skill: A Skill Collection That Turns Coding Agents Into Korean Life Assistants

> k-skill packages 125 skills for Korean everyday services, from KTX timetables to Coupang and government portals, and installs them into Claude Code, Codex, OpenCode and other agents with one npx command. The catch is that several skills go through a hosted proxy and some are read-only lookups rather than booking tools.

**NomaDamas/k-skill** — 한국인을 위한 스킬 모음집 - 에이전트를 한국인으로

- Repository: https://github.com/NomaDamas/k-skill
- Website: https://k-skill.nomadamas.org
- Stars: 7,727 · Forks: 940
- Language: JavaScript
- License: MIT
- Published: 2026-09-22 · Updated: 2026-09-22 · Language: en
- Canonical page: https://hysenlabs.com/projects/nomadamas-k-skill

## What k-skill actually covers, and who it is built for

The README opens with a direct pitch: if you are Korean, download this skill collection because you will definitely need it someday. The framing is not abstract. It lists KTX, KBO, Lotto, Daangn, Coupang, KakaoTalk, Government24 and HomeTax as the kind of chores you can hand to an AI agent, and the repository layout backs that up with 125 skills grouped into Korean daily-life categories: transport and travel, real estate and housing, law and public administration, business and tax, and more.

The target user is narrow and specific. You need to be running a coding agent that supports the skills format, and you need Korean services to be relevant to your work or life. The README names Claude Code, Codex, OpenCode and OpenClaw/ClawHub as supported clients, plus Vercel Agent Skills. If your agent is not on that list, the collection is not for you yet.

What makes the scope credible is the granularity. It is not one vague "Korea" skill. There is a separate skill for Seoul subway arrival times, one for Seoul bike rental stations, one for Seoul congestion levels across 121 hotspots, one for highway traffic and CCTV metadata, one for cheap gas stations near your current position, one for public parking lots, one for EV charger status. Each has its own docs page under docs/features/.

## How the skills reach Korean data sources

The architecture is deliberately thin. The README states that no additional client API layer is needed, and that when a skill does need a server, it can simply send an HTTP request to a proxy such as k-skill-proxy. That means the skills themselves are the integration point: they describe how to call a public API or read a public web surface, and the agent executes that.

The data sources fall into three rough groups. The first is official open APIs, often reached through the public proxy: the Ministry of Foreign Affairs travel advisory service, KOMSA ferry information, the National Tax Service business registration check, the National Pension workplace lookup, building register data from the Public Data Portal, and the National Assembly bill and vote API. The second is public web surfaces that are not documented OpenAPIs, and the README is honest about this: the housing official price skill reads realtyprice.kr public web data and is described as not being an officially documented OpenAPI. The third is third-party or commercial surfaces, such as the MyRealTrip MCP server, Kakao Local and Kakao Mobility, ODsay for transit routing, and fast-flights for Google Flights lookups.

That mix matters when you evaluate reliability. An official API with a proxy in front of it fails in predictable ways. A scraped public web surface fails whenever the page changes. The README does not document retry behaviour or rate limits for either case, so treat the distinction as a real operational difference rather than a documentation detail.

## Installing k-skill and running your first lookup

The default install needs only Node.js 18 or newer and npx. To install every skill globally, the README gives this command:

```bash
npx --yes skills add NomaDamas/k-skill --all -g
```

If you only want one, pass the skill name instead. The README's own example installs the railway timetable skill:

```bash
npx --yes skills add NomaDamas/k-skill --skill railway-timetable -g
```

The README also provides an update path for when the CLI itself is behind, which refreshes all coding agent skills including Vercel Agent Skills:

```bash
npx -y @nomadamas/k-skill@0 update
```

One dependency caveat is stated plainly. The KTX official timetable helper requires Python 3.11 or newer and uv in addition to Node.js. If you install only railway-timetable and skip Python, that helper will not run. Claude Code users have a second route through a marketplace, documented in docs/install.md.

After installing, the first real use is asking your agent for a railway timetable lookup. The README is explicit about the boundary of that skill: it queries Korail's official planned timetable and does not perform login, reservation, waiting-list booking, payment or cancellation. So the expected result is timetable data, not a ticket. Several skills that touch credentials or public data through the proxy are governed by the k-skill-proxy privacy policy, which the README links directly.

## Where k-skill stops short

The most important limitation is stated in a blockquote near the top of the README: the integrated railway timetable is a lookup-only skill. It reads Korail's official planned timetable and performs no login, reservation, waiting-list booking, payment or cancellation. Anyone who reads "KTX" and assumes ticket purchasing will be disappointed. The same read-only framing applies to the National Assembly bill and vote skill, which the README describes as lookup-only through the proxy.

A second limitation is the proxy dependency. Skills that use the public API proxy route your requests through infrastructure the project operates, and the README points to a separate privacy policy page for how that data is handled. If your use case involves anything sensitive, that is a decision you have to make explicitly rather than by default.

A third is data-source fragility. The housing official price skill and the individual land price skill both read realtyprice.kr public data, and the README labels the former as not an officially documented OpenAPI. Scraped surfaces break without notice, and the README does not document fallbacks.

Finally, some skills are explicitly assistive rather than autonomous. The IROS registry automation skill helps with the cart and the post-payment viewing and saving flow, and the court payment order assistant prepares drafts, checklists and a browser handoff after login. Those are human-in-the-loop by design, not a gap to be fixed.

## How k-skill differs from calling these APIs yourself

The obvious alternative is writing your own tool calls against the same public APIs. That is a real option, and for a single service it is often the better one: you control the error handling, the caching and the credentials, and you avoid the proxy entirely.

The difference in approach is that k-skill packages the endpoint knowledge, the parameter shapes and the output interpretation as agent-readable skills. The README's claim is that no additional client API layer is needed, so the work of figuring out which Public Data Portal dataset ID to use, what the region code format is, or how to express a subway station name is already done and maintained in the repository. The trade-off is that you inherit the project's choices about proxies, read-only boundaries and scraped surfaces.

A second alternative is a general-purpose MCP server for Korean services. The README notes that the MyRealTrip skill wraps an official MCP server for flights, accommodation and tours. That is a different model: one vendor's MCP surface versus a broad collection of individually scoped skills. If your needs are concentrated in travel booking, the official MCP route is narrower but deeper. If they are spread across tax, real estate, transit and retail, the breadth of k-skill is the point.

## Maintenance, licensing and what to check before adopting

The repository is not archived, and the last push was on 2026-09-21, one day before this writing. The most recent release listed is a rolling Manus bundle dated 2026-05-11. The project uses Changesets for versioning and publishing, with version-packages and release:npm scripts, and the CI script chains lint, typecheck, the main test suite, a store longevity mirror test and a pack dry run. That is a real pipeline, not a placeholder.

The licence is MIT, which is permissive and places few obligations on how you use the skills. One nuance worth noting without giving legal advice: the korean-privacy-terms skill is described as a thin wrapper around kimlawtech/korean-privacy-terms, which is Apache-2.0. If you redistribute or modify that component, a different licence applies to it than to the rest of the repository.

Before adopting, verify three things. First, read docs/install.md, because the README says per-agent installation details live there and the marketplace route for Claude Code is not spelled out in the main file. Second, check whether your target skills require Python 3.11 and uv, since that is an extra runtime beyond Node.js 18. Third, read the k-skill-proxy privacy policy if any skill you plan to use routes through the public proxy.

## Conclusion

Adopt k-skill if you already run a coding agent and want it to handle Korean public and commercial lookups: railway timetables, business registration checks, real estate transaction prices, congestion data. Do not adopt it expecting a booking engine, since the railway timetable skill is explicitly lookup-only and performs no login, reservation, payment or cancellation. Before installing, read docs/install.md for the per-agent setup paths, check the k-skill-proxy privacy policy if you plan to use any skill that routes through the public proxy, and confirm whether a skill you need requires Python 3.11 and uv on top of Node.js 18.

## FAQ

### What is NomaDamas/k-skill?

It is a collection of 125 skills that give coding agents the ability to handle Korean everyday services such as KTX timetables, KBO, Lotto, Daangn, Coupang, KakaoTalk, Government24 and HomeTax. The README describes it as a skill collection for Korean users, installable into Claude Code, Codex, OpenCode and OpenClaw/ClawHub.

### How do I install k-skill?

The README gives npx --yes skills add NomaDamas/k-skill --all -g to install everything, or the same command with --skill and a skill name to install one. Node.js 18 or newer and npx are the only base requirements, and the README provides npx -y @nomadamas/k-skill@0 update to refresh the CLI and all coding agent skills.

### Does the k-skill railway timetable skill book KTX tickets?

No. The README states in a blockquote that the integrated railway timetable is a lookup-only skill: it queries Korail's official planned timetable and does not perform login, reservation, waiting-list booking, payment or cancellation.

### Does k-skill need anything besides Node.js?

The base install needs Node.js 18 or newer and npx. The README adds that running the KTX official timetable helper requires Python 3.11 or newer and uv as well.

### What licence does k-skill use?

The repository is MIT licensed. The README notes that the korean-privacy-terms skill is a thin wrapper around kimlawtech/korean-privacy-terms, which is Apache-2.0, so that component carries a different licence.

## Sources

- [License: MIT](https://github.com/NomaDamas/k-skill/blob/main/LICENSE)
- [NomaDamas/k-skill on GitHub](https://github.com/NomaDamas/k-skill)
- [Project website](https://k-skill.nomadamas.org)
- [README](https://github.com/NomaDamas/k-skill/blob/main/README.md)
- [Releases](https://github.com/NomaDamas/k-skill/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/nomadamas-k-skill