Use when investigating any bug, error, crash, stack trace, failing or flaky test, incident, regression, unexpected behavior, or "why is this happening" question - before proposing or applying a fix.
Use when an implementation needs proof against the live running stack: after build completes, before ship, or when asked to verify, e2e-test, smoke-test, or demo that an API, UI, or DB change actually works.
Use when executing an implementation plan or building a scoped feature or fix - after scope is clear, before code gets written. Triggers on "implement the plan at PATH", "start slice SP-2b".
Use when reviewing code, a diff, a module, an architecture, or a design/spec document against house standards - including multi-expert panel reviews ("review with a DDD expert and an asset-finance expert"). Also when running a removal-first simplification sweep: "what can we remove", "product simplification review".
Use when building or styling sportsbook, betting, or casino (iGaming) interfaces: odds displays, betslips, live-event states, suspended markets, casino game grids, or wallet and money surfaces.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A 12-skill plugin encoding one engineer's actual workflow and house engineering law, rebuilt
2026-07-07 from usage evidence (2,597 real prompts) after the previous 48-skill generation
recorded zero invocations. The old pack lives at tag pre-redesign.
claude plugin marketplace add OmexIT/claude-skills-pack # or the local repo path
# in Claude Code: /plugin, enable "garage", restart session
codex plugin marketplace add OmexIT/claude-skills-pack # or the local repo path
codex plugin add garage@garage
# start a new Codex thread so the installed skills are loaded
Both clients install the same 12 skills, references, and skill scripts. The three lifecycle hooks use Claude Code's hook protocol and are loaded only by Claude Code; Codex's plugin manifest intentionally does not declare them.
Workflow spine (mirrors the real loop; each step optional):
| Skill | Use when |
|---|---|
spec | Defining or changing WHAT to build; PRD create/update; challenging scope |
plan | Turning an approved spec, ticket, or explicit request into a sliced, verifiable plan doc in docs/ |
build | Executing a plan ("implement docs/...plan.md", "start slice SP-2b") |
audit | Reviewing code, diffs, or docs against house standards; expert panels |
e2e | Live verification: boot the stack, newman, Playwright, DB checks. Evidence, not claims |
ship | Commit / PR / MR / branch promotion per each repo's delivery policy |
debug | Any bug or unexpected behavior, before proposing fixes |
Domain law (auto-triggers on subject matter):
| Skill | Carries |
|---|---|
ledger | Double-entry invariants, idempotency key grammar, Blnk + pgledger contracts |
migrations | TSID PKs, audit columns, RLS recipe, zero-downtime rules, throttled backfills |
temporal | SAGA compensation, retry profiles, versioning, failure-mode table |
spring-api | Response envelope, RFC 9457, package-by-feature layout, cursor pagination |
igaming-ui | Odds/betslip/live-state patterns, accessible selector contract |
Claude Code hooks (plugins/garage/hooks/hooks.json): destructive-command blocker,
sensitive-file warning, Prettier autoformat (web files only; Java/Spotless runs once in the
ship gate).
spec ──► plan ──► build ──► audit ──► e2e ──► ship
▲ │
└────────── debug ◄─────────┘
Plans live in each repo's docs/.../plans/ with checkboxes as durable session state.
Skills fire on natural phrasing. Force one explicitly with /garage:<skill> in Claude Code
or $garage:<skill> in Codex.
Single skills
"should we build bulk payouts, or does the batch API already cover it?" -> spec (challenge: keep/shrink/kill)
"write the PRD for payment links" -> spec (create)
"sync the PRD with what we actually built" -> spec (update, drift check)
"plan the wallet top-up feature from docs/prd/topups.md" -> plan (sliced plan doc)
"implement docs/plans/2026-07-08-topups-plan.md" -> build
"start slice 2b" -> build (resumes at that slice)
"review this diff against our standards" -> audit
"audit the payments module, include a DDD expert" -> audit (expert panel)
"simplification sweep: what can we remove from this product?" -> audit (removal-first, whole app)
"run and test orders end to end, api and ui" -> e2e (newman + Playwright + DB proof)
"why is the balance negative after a refund?" -> debug (diagnosis only, no fix yet)
"commit and merge to release-candidate" -> ship (repo's branch policy)
"promote develop to main" -> ship (promotion PR)
Domain skills load themselves when the work touches their subject: a hold/void flow pulls
ledger, a new column pulls migrations, a saga with compensation pulls temporal, a new
endpoint pulls spring-api, a betslip component pulls igaming-ui.
A feature, end to end
npx claudepluginhub omexit/claude-skills-pack --plugin garageSpec-driven development for big features. When features get too big, plan mode gets too vague—leading to hallucinations during implementation. ShipSpec replaces vague plans with structured PRDs, technical designs, and ordered tasks that keep Claude grounded.
Specification-driven development workflow: specify → plan → tasks → implement
Engineering process for solo founders and teams up to 50 engineers. Agents do architecture, code review, QA, and security. You make two decisions per feature.
PRD-driven AI dev team control plane for Claude Code — agents, skills, and a GitHub-first sprint workflow.
Helder's personal SDLC toolbelt for AI coding agents — from spec to ship. Bundles the tracer-bullet workflow alongside TDD, code review, audits, and shipping skills.
Verification-first engineering toolkit for Claude Code. 15 skills across a 5-phase spine (Investigate → Design → Implement → Verify → Ship), 8 specialist agents, an interactive setup wizard. Every skill has rationalizations + evidence requirements. Built for senior ICs and tech leads.