garage

Plugin garage from marketplace claude-garage in voyager67/Claude-Garage. Commands read from the repo's manifest — 8 days ago.

Skills (19)

• audit

  Adaptive parallel review before merge. Security, quality and tests are always reviewed; concurrency, dependencies, UI and infrastructure specialists join automatically when the diff calls for them. Returns one ranked verdict.

• blueprint

  Turn a pitch into a technical blueprint with trade-offs, interfaces, file-level changes and small build slices marked parallel-safe or not. For size L it also creates blind acceptance scenarios. Use after /pitch for M and L work.

• brainstorm

  Explore a fuzzy idea before committing to it. Generates conservative, ambitious and unconventional options, compares them and helps you pick one to pitch.

• break

  Adversarial testing. The gremlin agent attacks recent changes with hostile inputs, edge cases and bad state, and proves each finding with a failing test.

• docs

  Generate or synchronize documentation with the code. Creates README, API docs or architecture notes, and finds stale docs that no longer match reality.

• firefight

  Fix a bug or incident by root cause. Investigators work in parallel with competing hypotheses and try to disprove each other, then the surviving cause gets a minimal fix and a regression test.

• garage-init

  One-time setup. Creates the .garage workspace, seeds lessons.md, detects the test command, and offers house rules plus an optional output style.

• guide

  Interactive command finder. Asks what you want to do, checks the state of your .garage workspace, recommends the right command and offers to start it. Use whenever you are unsure what to run.

• launch

  Pre-release checklist and release writing. Verifies tests, docs and version, drafts the changelog and release notes, and never tags or pushes without confirmation.

• patterns

  Discover the conventions and recurring patterns already in a codebase and turn them into documented patterns and suggested house rules. Use to learn how a team really writes code.

• pitch

  Turn a rough idea into a one-page pitch with target user, MVP scope, non-goals, success metric, kill criteria and a size (S, M or L) that picks the lane. Use before writing code for any new feature or product.

• pr

  Prepare a pull request from your branch. Writes a clear title and description with the why, evidence, risks and a reviewer checklist. Never opens the PR without confirmation.

• preflight

  Quality gate for a pitch and blueprint before building. Scores them for completeness, consistency and correctness (testability), and optionally checks house-rule compliance. Advisory, never blocking.

• recon

  Map an unfamiliar codebase in one page, including structure, entry points, how to run and test, conventions and gotchas. Use when joining or inheriting a project.

• retro

  Short retrospective that turns what just happened into durable lessons. Writes actionable "when X, do Y" rules to .garage/lessons.md that every other command reads.

• rules

  Create or update the project's house rules (guardrails) in .garage/rules.md through a short interview. The rules are checked automatically by blueprint, sprint, preflight and audit.

• sprint

  Execute a pitch or blueprint. Auto-detects the lane (S, M or L), runs up to three builders in parallel on disjoint slices, checks for drift against the plan, runs blind scenarios for L work and parks scope creep.

• standup

  Instant status report from git history and .garage state. What shipped, what is in progress, what is blocked, and what to do next.

• tidy

  Behavior-preserving cleanup. Improves structure, naming and duplication in small steps with tests green after each step, and stops at the first red.

Subagents (13)

• architect

  Technical design. Use to turn a pitch into a blueprint with trade-offs, interfaces, file-level changes and small build slices.

• builder

  Implementation. Use to build exactly one blueprint slice with tests first and the smallest correct diff.

• chief

  Complexity assessment and routing. Use at the start of any piece of work to size it (S, M or L), pick the right lane and name which specialists should be involved.

• critic

  Code quality review. Use to check readability, duplication, naming, error handling and design drift against the blueprint.

• detective

  Root-cause debugging. Use when something is broken and the cause is not obvious. Forms competing hypotheses and tries to disprove them.

• gremlin

  Adversarial QA. Use to break new code with hostile inputs, edge cases, race conditions and bad state, and to prove each break with a failing test.

• pixel

  UX, interaction design and accessibility. Use for any user-facing change: flows, screens, copy, empty and error states, keyboard and screen-reader support.

• pm

  Product thinking. Use to sharpen scope, define the target user, write non-goals and success metrics, and cut features that do not earn their place.

• rigger

  Platform and delivery engineering. Use for CI/CD, containers, infrastructure as code, environment config, observability and anything that decides how the software runs in production.

• scout

  Read-only codebase recon. Use proactively to find what already exists, how things are wired, and where conventions live before anyone designs or edits anything.

• scribe

  Documentation and release writing. Use for READMEs, changelogs, release notes, docstrings and architecture notes that match the code as it really is.

• sentinel

  Security review. Use on any change touching auth, user input, secrets, file access, network calls or dependencies.

• tester

  Test strategy and blind acceptance scenarios. Use to decide what to test, to write scenarios builders never see, and to find coverage gaps.

Similar plugins