samber/cc-skills-golang

Worth a look if your coding assistant writes plausible but unidiomatic Go; install the full set rather than a couple of skills, because the authors designed them to reference each other.

The problem

A developer asks Claude Code, Codex or Cursor to add a handler, fix a race or review a package. The assistant produces something that compiles, yet it may ignore the project's preferred error wrapping, logging or testing structure, and the developer ends up rewriting it by hand or restating the same conventions in every prompt. With these skills installed, the assistant loads Go-specific guidance on demand when the task touches style, security, documentation or debugging, so the conventions arrive before the code does. The intended result is generated Go that needs less correction and fewer repeated instructions.

What's different

The mechanism is the Agent Skills format: reusable instruction sets that are loaded only when relevant, so they do not permanently occupy the context window. The repository is split into atomic units rather than one large prompt, and the units deliberately cross-reference one another — the README's own example puts error-handling rules that affect logging in golang-error-handling rather than in the observability skill.

The practical benefit is precision and scope: a performance task pulls in performance guidance, a security review pulls in audit material. The trade-off is coupling. The README warns that installing only a subset gives "a partial and potentially inconsistent view of the guidelines," so the advertised quality gain assumes near-complete installation, not cherry-picking. The README also states that the skills were bootstrapped with Claude Code by distilling the author's own Go commits and then edited and tested by a human; that provenance is a claim, and the evaluation below is the author's own evidence rather than an independent benchmark.

Who should try it

Good fit: a Go team that already uses an Agent Skills-compatible assistant and keeps losing time to the same style, error-handling and debugging corrections. The all-at-once installation and the cross-references make sense for a team adopting a shared convention baseline.

Poor fit: projects with their own strongly enforced lint and review rules that would conflict rather than be complemented, and users whose assistant is not compatible with the Agent Skills format — for those, the README points to the universal skills CLI or the Claude Code plugin marketplace as install paths, but compatibility still depends on the client. The repository covers Go only; general workflow skills such as git conventions, CI/CD and PR review are deliberately left out and live in a separate plugin.

Requirements and limitations

You need an assistant that can load Agent Skills; the README lists Claude Code, Codex, Gemini, Copilot, Antigravity and Cursor, with the skills CLI as the universal route and an Openclaw path that copies files into a discovery directory. Installing the whole set is the documented setup, so expect a fair amount of guidance to be added and plan how your team's own conventions override it — several skills are described as overridable by company skills.

Ongoing upkeep matters: these are rules about a moving language and ecosystem, and they were distilled from one developer's projects. Whether they match your codebase's idioms is something your review process has to check; installing them is not a substitute for lint, tests or a human reviewer.

The strongest evidence supplied is the author's evaluation table showing 3348/3439 (97%) with skills versus 1957/3439 (57%) without, a +40pp delta, with per-skill detail in EVALUATIONS.md. Treat that as a vendor-published result, not independent verification.

HotRepo's take

The repository is 219 days old and was seen at #8 on the weekly Go board with +171 stars for the week on 2026-10-27. That is a single recorded weekly observation, not a daily series, and HotRepo has no daily-board history, so persistence of interest cannot be judged; there is also a March 2026 HN thread with 3 points and 2 comments, which indicates discussion volume only. Before adopting, check the per-skill evaluation breakdown and confirm that the skills' conventions match your project's existing Go style rules, since the full-set design assumes you accept most of them.

Project source and documentation