Use when building Grape APIs: endpoint design, param coercion, declared params, versioning, error handling, grape-entity presentation, and Rails integration.

Use when building native iOS/Android mobile apps with Hotwire Native and Turbo Native. Covers path configuration, bridge components, native screens, and web-to-native communication.

Use when checking or enforcing the 22 Iron Laws of Ruby/Rails/Grape development. Non-negotiable rules that prevent common, costly mistakes.

Use when building Kafka consumers or producers with Karafka in Ruby. Covers multi-threaded consumers, batch processing, error handling, and Rails integration.

Use when applying idiomatic Rails patterns for controllers, params, callbacks, jobs, routes, and application structure across Rails 7.x and 8.x.

Use when you need a broad project-wide audit of a Ruby/Rails/Grape codebase covering architecture, security, performance, testing, and operational risk.

Use when a user wants a concise explanation of a plan, implementation, or review outcome. Use after /rb:plan, /rb:work, or /rb:review for a high-signal walkthrough.

Use when you want rigorous adversarial scrutiny of Active Record changes, Hotwire/Turbo events, or PR readiness -- challenge mode questioning before approving.

Use when sharing verify-output compression telemetry with the plugin maintainers as an anonymized GitHub-issue-ready report. Reads the compression jsonl in the plugin data directory, redacts PII, drafts an actionable report with sample-by-sample reasoning, hands it back for the user to copy into a GitHub issue.

Use when diagnosing Active Record constraint failures, unique index violations, foreign-key errors, and migration/data mismatches.

Drafting an internal report from skill-discovery telemetry collected when RUBY_PLUGIN_DISCOVERY_LOG=1 is set. Reads ${CLAUDE_PLUGIN_DATA}/discovery.jsonl, summarizes hook events that matched trigger patterns, lists candidate-skill suggestions, computes per-rule match rate, suggests trigger-table tweaks. Use to evaluate whether to flip the layer from log-only to actively-injecting in a future release.

Use when generating documentation for implemented features — YARD/RDoc, README updates, ADRs. Use after implementing new modules or features to ensure proper documentation coverage. Run after /rb:review passes or whenever the user asks to document their code.

Use when the user wants to see example areas and representative patterns inside the current Ruby/Rails/Grape codebase.

Use when initializing the Ruby/Rails/Grape plugin in a project. Installs auto-activation rules into CLAUDE.md for complexity detection, workflow routing, Iron Laws, and Ruby stack auto-loading.

Use when users want a tutorial walkthrough of plugin commands and capabilities. Use when users want to learn what the plugin offers, not when they already have a specific task to work on.

Use when something is broken and the cause is unknown -- root-cause analysis for bugs, errors, and failures. NOT for performance tuning (use /rb:perf) or code flow tracing (use /rb:trace).

Use when diagnosing and explaining N+1 query patterns in Rails and Grape codepaths. Use for slow index pages, serializers, nested API responses, and Hotwire screens pulling associated data.

Use when analyzing performance issues in Rails, Grape, Active Record, Redis, and Sidekiq. Use for slow endpoints, queue latency, cache misses, object churn, or N+1 reports.

Use when you need an implementation plan for multi-file Rails or Grape features, Sidekiq changes, or risky migrations and refactors before coding starts. Also accepts review files and existing plans.

Use when auditing existing research/review `*.provenance.md` sidecars under `.claude/`. Classifies each by `trust_state` (clean / weak / conflicted / missing) and writes a dated Markdown report of the distribution and per-file breakdown.