# Ruby Engineering > Ruby engineering guidance. Use when creating, changing, reviewing, testing, packaging, or scripting Ruby source, gems, Gemfiles, Bundler, Rake, Minitest, RSpec, Rails-adjacent Ruby, or Ruby CLIs. Do not use merely to choose among scripting languages without Ruby implementation, or for framework-specific behavior whose owning skill is more precise. - Skill: `nledford/ruby-engineering` (Agent Skill) - Install (CLI): `npx skillmds@latest add nledford/ruby-engineering` - Raw SKILL.md: https://api.skillmd.com/api/skills/nledford/ruby-engineering/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: nledford (https://skillmd.com/u/nledford) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/nledford/ruby-engineering --- # Ruby Engineering Use this skill for repository-backed Ruby implementation and workflow decisions. Inspect the supported Ruby version, Bundler and gem policy, source layout, test framework, style tools, and execution targets before assuming a universal Ruby setup. Use [`script-engineering`](../script-engineering/SKILL.md) when the primary decision is whether automation should be a script or whether Ruby fits better than shell, Python, PowerShell, or JavaScript. This skill owns the Ruby mechanics after Ruby is selected. ## MCP SDK Deferral For an MCP request, load [`mcp-server-engineering`](../mcp-server-engineering/SKILL.md) and its dated [SDK selection record](../mcp-server-engineering/references/sdk-selection.md). That record owns Ruby's current eligibility, authority, package/version, tier, core-revision, transport, lifecycle, and conformance caveats. Do not prescribe Ruby MCP gem APIs or server/client wiring from this skill. Defer implementation selection until current official evidence and target validation qualify the selected surface; retain this skill for ordinary Ruby mechanics only. ## Workflow 1. Read repository instructions, `.ruby-version`, version-manager files, `Gemfile`, `Gemfile.lock`, gemspecs, `Rakefile`, source/tests, CI, and existing Ruby commands. 2. Identify supported Ruby and Bundler versions, target operating systems, and whether the code is an application, library, gem, CLI, or repository script. 3. Preserve the repository's framework and conventions. Do not introduce Rails, RSpec, Minitest, RuboCop, Sorbet, Steep, or another tool without evidence or an explicit adoption goal. 4. Define observable behavior and choose the narrowest useful test boundary. 5. Implement small explicit objects, modules, and functions; isolate external processes, filesystems, clocks, networks, and databases behind testable boundaries. 6. Run focused syntax, test, lint, and packaging checks before broader repository-native validation. 7. Report versions and commands observed, files and dependencies changed, validation run, skipped targets, and remaining compatibility risk. ## Runtime And Dependency Discipline - Treat `.ruby-version`, `.tool-versions`, `mise.toml`, gem metadata, CI images, and deployment configuration as runtime evidence. A locally available `ruby` does not prove every target supports that version. - Use the repository's Bundler workflow. Prefer `bundle exec ` for project-owned executables when that is the established way to bind commands to the lockfile. - Keep `Gemfile` intent and `Gemfile.lock` changes aligned. Do not hand-edit lockfile resolution or introduce a global gem dependency for project behavior. - Prefer the standard library for small needs when it is clear and sufficient. Add a gem only when its maintained behavior outweighs dependency, native extension, startup, licensing, and supply-chain cost. - Load [`dependency-supply-chain-review`](../dependency-supply-chain-review/SKILL.md) for new gems, lockfile churn, native extensions, sources, install hooks, provenance, or advisory questions. ## Local CPU Parallelism Routing Compose with [`parallelism-engineering`](../parallelism-engineering/SKILL.md) when a local CPU-bound Ruby workload needs data or task decomposition, partition sizing, bounded worker-pool design, deterministic reduction, cancellation, or nested-parallelism control. That skill owns the cross-language parallel design; this skill owns Ruby runtime, library, gem, test, and process/thread mechanics. Do not route routine async I/O, event-loop behavior, fibers, ordinary background work, or generic Ruby concurrency to `parallelism-engineering`; keep that work with Ruby's implementation owner. Spark/PySpark execution remains with [`data-platform-engineering`](../data-platform-engineering/SKILL.md), not local parallelism engineering. ## WebAssembly Routing Compose with [`webassembly-engineering`](../webassembly-engineering/SKILL.md) when Ruby work includes a general WebAssembly decision: a WAT or `.wasm` artifact, WASI or WIT, the Component Model, host/guest contract, runtime or target selection, capability grants, or Wasm packaging and deployment. That skill owns those Wasm boundary and compatibility decisions. This skill retains Ruby source, gem, compiler or binding-tool, host-integration, and test mechanics; the Wasm skill does not establish Ruby toolchain or binding support. ## Idiomatic Ruby - Prefer clear message-oriented objects, modules, blocks, enumerators, keyword arguments, and standard collection APIs over manual indexing or framework ceremony. - Keep methods cohesive and make mutation visible. Return useful domain values instead of relying on unrelated global or process state. - Use modules for shared behavior or namespaces deliberately; do not use mixins as an unstructured substitute for object boundaries. - Make public APIs explicit about accepted values, returned values, exceptions, side effects, and destructive or mutating behavior. Use `?` and `!` names only when they follow established Ruby meaning in the repository. - Prefer ordinary methods, delegation, and explicit registration over dynamic dispatch. Use metaprogramming only when it materially reduces repeated structure and leaves discoverable behavior, tests, and error messages. - Preserve exception context. Rescue only errors the current boundary can handle, and convert them into a stable domain or CLI result where appropriate. - Keep load-time work cheap and deterministic. Avoid network, filesystem, process, or environment-dependent side effects during `require`. - Follow the repository's formatter and style configuration instead of applying a different community style mechanically. ## Ruby Anti-Patterns Avoid: - global monkey patches to core classes or dependencies; - `method_missing`, callbacks, DSL magic, or reflection where an explicit API is simpler and easier to test; - mutable constants, global variables, class variables used as shared state, or singleton service registries that hide ownership; - rescuing `Exception`, broad rescue-and-continue behavior, swallowed failures, or retry loops without bounds and an error policy; - shell command strings assembled from input, backticks for complex process control, or ignored subprocess exit status; - implicit working-directory assumptions and unscoped temporary files; - tests that stub the object under test, assert private call order, or replace so many collaborators that no real behavior remains; and - framework models, callbacks, or persistence types leaking into domain APIs without an intentional boundary. ## Testing Ruby Use the test framework already owned by the repository. - **Unit:** test value objects, parsers, formatters, validation, domain rules, enumerators, and error behavior without external resources. - **Integration:** exercise real filesystem, subprocess, database, framework, gem, serialization, and adapter boundaries in isolated fixtures. - **CLI/process:** invoke the executable when arguments, environment, stdout, stderr, signals, or exit status are part of the contract. - **End to end:** cover a complete supported user or operator workflow only when lower levels cannot prove the behavior. Use the repository's web/system-test stack; do not assume Rails, Capybara, Selenium, or a browser runner. For Minitest, preserve the repository's spec-style or class-style convention and run its configured Rake or direct Ruby lane. For RSpec, preserve configured formatters, tags, helpers, and suite boundaries and use the Bundler-bound command. Do not migrate between them merely for preference. Keep tests deterministic and isolated. Use temporary directories, fake clocks, seeded randomness, and boundary substitutes where needed; avoid live networks, shared databases, process-global leakage, order dependence, and unbounded sleeps. Mock external boundaries, not the behavior being specified. Typical checks to derive from repository evidence include: ```sh ruby -c path/to/file.rb bundle exec ruby -Itest path/to/test_file.rb bundle exec rake test bundle exec rspec path/to/spec_file.rb bundle exec rubocop ``` Do not claim a command exists or passed without observing it. Use [`test-driven-development`](../test-driven-development/SKILL.md) for behavior changes and regressions, and [`systematic-debugging`](../systematic-debugging/SKILL.md) for an active unexplained Ruby failure. ## Ruby As A Script Choice Prefer Ruby when all required targets provide the supported runtime and one or more of these are true: - the repository already uses Ruby and the script can reuse maintained code or locked gems; - structured text, files, APIs, reusable functions, or CLI parsing have outgrown a small shell wrapper; - Ruby's collection, string, regular-expression, or DSL capabilities express the task clearly without introducing an additional runtime; or - the automation is naturally adjacent to a Ruby application, gem, or Rake workflow. Do not choose Ruby when a short portable shell command is clearer, Ruby is absent from a bootstrap or deployment target, the repository already has a suitable supported automation language, or author familiarity is the only advantage. Domain behavior that callers should import belongs in the application or library, with a thin script or CLI entry point at most. ## Security And Completion Load [`security-review`](../security-review/SKILL.md) when Ruby code handles untrusted input, paths, serialization, templates, commands, credentials, privileged operations, or remote targets. Use [`security-review-evidence`](../security-review-evidence/SKILL.md) when logs, fixtures, reports, or reproduced output may contain sensitive values. Before handoff, report the Ruby and Bundler evidence, why Ruby fit the task, tests and quality gates run, dependency or lockfile impact, skipped platforms, and residual runtime, compatibility, or security risk.