What problem does it solve? Go code written or reviewed by AI assistants often reads as machine-generated — essayistic comments, reflexive doc blocks, and uniform density — and misses the specific conventions the Miren runtime team enforces in review. This Skill encodes the team's actual house style, derived from the mirendev/runtime codebase and roughly 1,600 real review comments by maintainers phinze and evanphx, so contributions look Miren-authored and reviews flag what maintainers actually flag. ## Core Features & Use Cases - Writing mode: Eliminates LLM "tells" (free-floating file-overview prose, essay comments, _ struct{} opts) and applies universal conventions for command handlers, error wrapping with fmt.Errorf("...: %w", err), naming, table-driven tests, concurrency, and import layout. - Reviewing mode: Provides the substantive adjustment catalog maintainers use — reuse over reinvention, scoped error handling, concurrency bounds, lifecycle tracing, security, scope discipline — plus the blocking calculus and suggestion-reply taxonomy. - Sanctioned inconsistencies: Distinguishes universal conventions (flag them) from sanctioned mixed styles like ID vs Id or testify vs plain t.Errorf (never flag them). - Use Case: When reviewing a Miren PR that adds a new CLI command, verify the handler signature func Name(ctx *Context, opts struct{...}) error, check for duplicated helpers, and label each finding blocking or not — exactly as phinze or evanphx would. ## Quick Start Ask the assistant to review this Go pull request against the Miren house style and list which findings are blocking.
