Go Code Review Agent Skill

Go Code Review Checklist

> Compatibility: references/WEB-SERVER.md uses log/slog examples that require Go 1.21+.

Resource Routing

• assets/review-template.md - Use when formatting review output with Must Fix, Should Fix, and Nits sections.
• scripts/pre-review.sh - Run before manual review to collect gofmt, go vet, and golangci-lint results.
• references/WEB-SERVER.md - Read when reviewing an HTTP server that combines concurrency, context, logging, error handling, and shutdown behavior.

Review Procedure

> Use assets/review-template.md when formatting the output of a code review to ensure consistent structure with Must Fix / Should Fix / Nits severity grouping.

• Run gofmt -d . and go vet ./... to catch mechanical issues first
• Read the diff file-by-file; for each file, check the categories below in order
• Flag issues with specific line references and the rule name
• After reviewing all files, re-read flagged items to verify they're genuine issues
• Summarize findings grouped by severity (must-fix, should-fix, nit)

> Validation: After completing the review, re-read the diff once more to verify every flagged issue is real. Remove any finding you cannot justify with a specific line reference.

Formatting

• ] gofmt: Code is formatted with gofmt or goimports → [go-linting

Documentation

• ] Comment sentences: Comments are full sentences starting with the name being described, ending with a period → [go-documentation
• ] Doc comments: All exported names have doc comments; non-trivial unexported declarations too → [go-documentation
• ] Package comments: Package comment appears adjacent to package clause with no blank line → [go-documentation
• ] Named result parameters: Only used when they clarify meaning (e.g., multiple same-type returns), not just to enable naked returns → [go-documentation

Error Handling

• ] Handle errors: No discarded errors with _; handle, return, or (exceptionally) panic → [go-error-handling
• ] Error strings: Lowercase, no punctuation (unless starting with proper noun/acronym) → [go-error-handling
• ] In-band errors: No magic values (-1, "", nil); use multiple returns with error or ok bool → [go-error-handling
• ] Indent error flow: Handle errors first and return; keep normal path at minimal indentation → [go-error-handling

Naming

• ] MixedCaps: Use MixedCaps or mixedCaps, never underscores; unexported is maxLength not MAX_LENGTH → [go-naming
• ] Initialisms: Keep consistent case: URL/url, ID/id, HTTP/http (e.g., ServeHTTP, xmlHTTPRequest) → [go-naming
• ] Variable names: Short names for limited scope (i, r, c); longer names for wider scope → [go-naming
• ] Receiver names: One or two letter abbreviation of type (c for Client); no this, self, me; consistent across methods → [go-naming
• ] Package names: No stuttering (use chubby.File not chubby.ChubbyFile); avoid util, common, misc → [go-packages
• ] Avoid built-in names: Don't shadow error, string, len, cap, append, copy, new, make → [go-declarations

Concurrency

• ] Goroutine lifetimes: Clear when/whether goroutines exit; document if not obvious → [go-concurrency
• ] Synchronous functions: Prefer sync over async; let callers add concurrency if needed → [go-concurrency
• ] Contexts: First parameter; not in structs; no custom Context types; pass even if you think you don't need to → [go-context

Interfaces

• ] Interface location: Define in consumer package, not implementor; return concrete types from producers → [go-interfaces
• ] No premature interfaces: Don't define before used; don't define "for mocking" on implementor side → [go-interfaces
• ] Receiver type: Use pointer if mutating, has sync fields, or is large; value for small immutable types; don't mix → [go-interfaces

Data Structures

• ] Empty slices: Prefer var t []string (nil) over t := []string{} (non-nil zero-length) → [go-data-structures
• ] Copying: Be careful copying structs with pointer/slice fields; don't copy *T methods' receivers by value → [go-data-structures

Security

• ] Crypto rand: Use crypto/rand for keys, not math/rand → [go-defensive
• ] Don't panic: Use error returns for normal error handling; panic only for truly exceptional cases → [go-defensive

Declarations and Initialization

• ] Group similar: Related var/const/type in parenthesized blocks; separate unrelated → [go-declarations
• ] var vs :=: Use var for intentional zero values; := for explicit assignments → [go-declarations
• ] Reduce scope: Move declarations close to usage; use if-init to limit variable scope → [go-declarations
• ] Struct init: Always use field names; omit zero fields; var for zero structs → [go-declarations
• ] Use any: Prefer any over interface{} in new code → [go-declarations

Functions

• ] File ordering: Types → constructors → exported methods → unexported → utilities → [go-functions
• ] Signature formatting: All args on own lines with trailing comma when wrapping → [go-functions
• ] Naked parameters: Add /* name */ comments for ambiguous bool/int args, or use custom types → [go-functions
• ] Printf naming: Functions accepting format strings end in f for go vet → [go-functions

Style

• ] Line length: No rigid limit, but avoid uncomfortably long lines; break by semantics, not arbitrary length → [go-style-core
• ] Naked returns: Only in short functions; explicit returns in medium/large functions → [go-style-core
• ] Pass values: Don't use pointers just to save bytes; pass string not *string for small fixed-size types → [go-performance
• ] String concatenation: + for simple; fmt.Sprintf for formatting; strings.Builder for loops → [go-performance

Logging

• ] Use slog: New code uses log/slog, not log or fmt.Println for operational logging → [go-logging
• ] Structured fields: Log messages use static strings with key-value attributes, not fmt.Sprintf → [go-logging
• ] Appropriate levels: Debug for developer tracing, Info for notable events, Warn for recoverable issues, Error for failures → [go-logging
• ] No secrets in logs: PII, credentials, and tokens are never logged → [go-logging

Imports

• ] Import groups: Standard library first, then blank line, then external packages → [go-packages
• ] Import renaming: Avoid unless collision; rename local/project-specific import on collision → [go-packages
• ] Import blank: import _ "pkg" only in main package or tests → [go-packages
• ] Import dot: Only for circular dependency workarounds in tests → [go-packages

Generics

• ] When to use: Only when multiple types share identical logic and interfaces don't suffice → [go-generics
• ] Type aliases: Use definitions for new types; aliases only for package migration → [go-generics

Testing

• ] Examples: Include runnable Example functions or tests demonstrating usage → [go-documentation
• ] Useful test failures: Messages include what was wrong, inputs, got, and want; order is got != want → [go-testing
• ] TestMain: Use only when all tests need common setup with teardown; prefer scoped helpers first → [go-testing
• ] Real transports: Prefer httptest.NewServer + real client over mocking HTTP → [go-testing

Automated Checks

Run automated pre-review checks:

bash scripts/pre-review.sh ./...         # text output
bash scripts/pre-review.sh --json ./...  # structured JSON output

Or manually: gofmt -l <path> && go vet ./... && golangci-lint run ./...

Fix any issues before proceeding to the checklist above. For linter setup and configuration, see go-linting.

Related Skills

• Style foundations: See go-style-core when resolving formatting debates or applying the clarity > simplicity > concision priority
• Linting setup: See go-linting when configuring golangci-lint or adding automated checks to CI
• Error strategy: See go-error-handling when reviewing error wrapping, sentinel errors, or the handle-once pattern
• Naming conventions: See go-naming when evaluating identifier names, receiver names, or package-symbol stuttering
• Testing patterns: See go-testing when reviewing test code for table-driven structure, failure messages, or helper usage
• Concurrency safety: See go-concurrency when reviewing goroutine lifetimes, channel usage, or mutex placement
• Logging practices: See go-logging when reviewing log usage, structured logging, or slog configuration