vinvcn/mattpocock-skills-zh-CN: a Simplified Chinese localisation of Matt Pocock's agent skills
这是 mattpocock/skills 的简体中文本地化版本。
At a glance
- What is it?
- This repository keeps the directory names, skill names, commands and tool identifiers of mattpocock/skills untouched while translating the prose into Simplified Chinese. It installs through the same skills.sh installer, but the sync model and the plugin route are worth understanding before you commit a project to it.
- Who is it for?
- Adopt this repository if your team works primarily in Chinese and you want the grilling, TDD and bug-diagnosis skills readable without constant code-switching; the translated prose is the whole point, and the install path is the same one-liner as upstream.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 2 days ago.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.
Editorial analysis
What the Chinese localisation actually changes, and what it refuses to touch
The repository is a Simplified Chinese localisation of mattpocock/skills. The README is explicit about the boundary: documentation and skill instructions are translated, while directory names, skill names, commands, code blocks, paths and tool identifiers stay as they are so that installation and runtime behaviour do not break. That constraint is the interesting design decision here. A skill is loaded by name and lives at a known path, so translating the name would break invocation; translating only the prose keeps the machinery intact while changing what the model reads.
The stated audience is Chinese-native developers and models whose main interaction language, or training corpus, is Chinese. The README argues this is not only about reading comfort: Chinese prompts and skill instructions fit a Chinese context better and reduce ambiguity from mixed Chinese and English. That is a claim about model behaviour, not a measured result, and the repository offers no benchmark for it. Treat the benefit as plausible but unquantified.
The translation strategy is described as skill-guided content localization, with the upstream repository treated as the English source. Synchronised translation is performed by Codex and merged into main by the maintainer through pull requests. The maintenance rules live in .skills/translate-skill/SKILL.md, which is the file to read if you want to know exactly which strings are considered behaviour-critical.
How the sync model works, and why Git history is deliberately not mirrored
The README states that this repository synchronises with upstream by content refresh, and does not sync upstream Git history or repository management state. That is a meaningful divergence from a plain fork. A fork carries history and lets you diff against upstream; this repository carries translated content and a translation workflow, so your ability to answer "what changed upstream and when" from this repository alone is limited. SYNC.md and the scripts directory sit at the top level, which suggests the refresh is at least partly scripted, but the README does not document a rollback procedure for a bad sync.
The user-visible install path is pinned to vinvcn/mattpocock-skills-zh-CN rather than upstream, so anything that resolves the repository by name resolves to this one. Skill names inside remain the upstream ones. That split is what makes the localisation usable: your agent sees a skill called the same thing it would see upstream, and the human-readable instructions around it are in Chinese.
One consequence deserves attention. Because history is not mirrored, the version tags here (v1.2.3 and v1.1.0 are the two most recent releases) reflect this repository's release cadence, not upstream's. If you are pinning for reproducibility, pin to a tag here and read the corresponding upstream content, rather than assuming the tag numbers line up.
Installing the Chinese skills and running the first setup pass
The README gives a 30-second install. The installer is the same skills.sh command used upstream, pointed at this repository. It copies skills into your project and asks which skills and which coding agents you want.
npx skills@latest add vinvcn/mattpocock-skills-zh-CNDuring the first install the README says to make sure you select /setup-matt-pocock-skills. After the copy finishes, run that skill inside your agent. According to the README it asks three things: which issue tracker you use (GitHub, Linear or local files), which labels you use when triaging issues (the /triage skill reads those labels), and where created docs should be saved. The setup pass is what wires the rest of the skills to your repository, so skipping it leaves skills that reference a tracker and a docs directory you never configured.
If you prefer a managed bundle over editable files, the README offers a Claude Code plugin route instead.
/plugin marketplace add vinvcn/mattpocock-skills-zh-CN
/plugin install mattpocock-skills@mattpocockThe README also documents a shell equivalent using the claude CLI, and a DeepSeek Harness install as a plugin:
dsh plugin --profile web add @vinvcn/dsh-mattpocock-skills-zhAfter installation, skills are invoked with a /zh- prefix, for example /zh-grilling. The README does not list every available skill in the section quoted here; the skills directory is the place to look.
Copy-in install versus managed plugin: the duplication trap
The README frames the two install routes as two different postures, and says to pick one. Installing both causes every skill to be installed twice. The skills.sh route copies skills into the project so you can edit and customise them. The plugin route installs them as a managed, read-only bundle that updates as new versions are released, which the README describes as a subscription rather than a fork.
That trade-off is real and it is the decision most teams will get wrong. If you intend to modify skill instructions for your own conventions, the copy-in route is the only one that supports it, and you take on the job of re-applying your edits after each content refresh. If you want to follow upstream changes without maintaining anything, the plugin route does that, at the cost of losing the ability to edit. There is no documented middle path in the README, and no documented merge strategy for local edits across a sync.
A second constraint: the README states that a native Codex plugin is not currently provided, though the skills.sh installer can install these skills into Codex and other Agent Skills-compatible harnesses. So Codex users get the copy-in route whether or not they wanted the managed one.
What the skills are for: grilling, shared language, TDD and bug diagnosis
The upstream README, reproduced in translation here, frames the skills as a response to four recurring failure modes with coding agents. The first is misalignment: the agent builds something other than what you meant. The answer is a grilling session, available as /grill-me for non-code situations and /grill-with-docs for engineering work, which asks detailed questions about the change before it starts.
The second is verbosity, traced to agents lacking a shared vocabulary for the project. /grill-with-docs builds a shared language document and records hard-to-explain decisions as ADRs. The README claims this reduces token spend and makes the codebase easier for an agent to navigate, because variable, function and file names converge on the same terms. The example given is a CONTEXT.md from the author's course-video-manager repository, where a long sentence about lessons and file system positions collapses into a short phrase about a materialization cascade.
The third is code that does not run, addressed through feedback loops: static types, browser access and automated tests, with /tdd enforcing a red-green-refactor cycle and /diagnosing-bugs wrapping debugging practice into a staged loop. The fourth is design decay, which the README addresses with the same design-every-day framing. Note that the quoted README is truncated during the fourth point, so the design-related skills are not fully enumerated in the available text.
Where this localisation is the wrong choice
The clearest limitation is the one the README itself states: there is no native Codex plugin. If your workflow depends on managed plugin packaging in Codex, this repository does not provide it, and you fall back to the installer.
The second is the history gap. Content refresh without Git history means you cannot use this repository as your audit trail for upstream changes. If your team needs to review exactly what changed in a skill between two upstream revisions, this repository is the wrong primary source; upstream is.
Third, translation introduces a maintenance lag by construction. New upstream prose has to be translated and merged before it appears here. The README does not state a sync cadence, so there is no documented guarantee about how far behind upstream content can fall. For a repository whose skills shape how an agent plans a change, that lag is a real operational property, not a cosmetic one.
Finally, the localisation changes only natural-language instructions. If your difficulty with these skills is structural (the workflow does not match how your team ships), Chinese prose will not fix it, and you will have taken on a sync dependency for nothing.
Licence and the cost of keeping up
The repository is MIT licensed, and it ships both LICENSE and LICENSE.zh-CN.md at the top level. MIT is permissive, so modification and redistribution are contemplated by the licence itself, which fits the copy-in install route where you edit skills in place. The translated licence file existing alongside the original is a sensible touch for a localisation repository. This is a description of what the repository contains, not legal advice; if licence obligations matter to your organisation, read both files and the upstream licence.
Upgrade cost depends on which route you chose. With the managed plugin, the README's framing is that new versions update as a bundle, so the cost is following releases. With the copy-in route, the cost is re-applying your local edits after each refresh, and the README does not document a merge or rollback procedure for that. The two most recent releases are v1.2.3 and v1.1.0, and the last push to the repository was on 2026-09-21. The README does not publish a support window or a compatibility matrix for agent versions, so verify against your own agent before rolling it out broadly.
Editorial conclusion
Adopt this repository if your team works primarily in Chinese and you want the grilling, TDD and bug-diagnosis skills readable without constant code-switching; the translated prose is the whole point, and the install path is the same one-liner as upstream. Do not adopt it if you need native Codex plugin packaging, since the README states that no native Codex plugin is provided, or if you intend to track upstream Git history, because this repository syncs content rather than history. Before installing, read .skills/translate-skill/SKILL.md to see how translations are produced and what stays in English, and decide between the skills.sh copy-into-project route and the managed Claude Code plugin, because the README warns that installing both duplicates every skill.
Frequently asked questions
What are Matt Pocock's skills?
They are a set of small, composable agent skills for real engineering work, covering alignment sessions, shared project language, test-driven development and bug diagnosis. This repository is the Simplified Chinese localisation of that collection, with prose translated and skill names, commands and paths left unchanged.
What is the Matt Pocock grilling skill?
It is a session in which the agent asks detailed questions about what you intend to build, so you catch misalignment before code is written. The README lists /grill-me for non-code situations and /grill-with-docs for engineering work, the latter also building shared language and recording decisions as ADRs.
How do I update Matt Pocock's skills?
It depends on how you installed them. The README says the Claude Code plugin installs the skills as a managed bundle that updates as new versions are released, while the skills.sh route copies files into your project, so you re-run the installer or re-apply your local edits. The README does not document a rollback procedure for a sync.
How do I set up Matt Pocock skills?
Run npx skills@latest add vinvcn/mattpocock-skills-zh-CN, choose the skills and coding agents, and make sure to select /setup-matt-pocock-skills during the first install. Then run that skill in your agent, which asks which issue tracker you use, which triage labels you use, and where docs should be saved.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/vinvcn-mattpocock-skills-zh-cn)