Go Guidelines

Go Guidelines

Make your AI code agent write production-grade Go instead of tutorial-grade Go.

A composable skill plugin for coding agents — covering modern syntax, generics, performance, concurrency safety, error handling, testing, and post-change verification including go fix on Go 1.26+.

Quickstart

Install Go Guidelines for your agent: Claude Code, Codex CLI, Cursor, OpenCode, Pi.

Motivation

All coding agents tend to generate outdated and suboptimal Go. Key reasons:

1. Training data lag. Models don't know about features added after their training cutoff. They can't use wg.Go() (1.25), new(val) (1.26), or generic methods (1.27) if they've never seen them.

2. Frequency bias. Even for features the model knows, it picks older patterns. There's more for i := 0; i < n; i++ in the training data than for i := range n, so that's what comes out.

3. No performance awareness. Agents don't align struct fields, don't set GOMAXPROCS for containers, and spawn unbounded goroutines instead of using pools.

4. Broken operational patterns. Generated shutdown code closes the database before draining the HTTP server. Resources get leaked. Signals get ignored.

5. No post-change verification. Agents never run golangci-lint, go fix (Go 1.26+), or tests with the race detector after making changes.

These guidelines fix all of the above by giving the agent an explicit, version-aware reference. The agent detects your Go version from go.mod and applies only the features and patterns available up to that version.

This aligns with the Go team's direction. The rewritten go fix command (Go 1.26+) modernizes existing code with analyzers for newer idioms. These guidelines serve the same goal for *new* code — and on Go 1.26+ they also require agents to run go fix after changes.

How it…