engineering

Plugin engineering from marketplace xxkeefer-skills in xxKeefer/skills. Commands read from the repo's manifest — today.

Skills (77)

• apply-review

  Apply the rows that you tick in a review report of your own branch. Fix test-first, commit once, push once, then reply on and resolve the ticked threads.

• claim

  Find or create the worktree for a ticket key, a PR, or a branch, then record the branch and the worktree in the handoff for that work. Needs worktrunk (`wt`). Use when the user says "claim" or asks which worktree holds a ticket, or when another skill needs a worktree before work starts.

• code-simplify

  ★ Code-simplify: make working code easier to read without a change in behaviour, one change at a time, with the tests run after each change.

• codebase-design

  Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.

• codebase-news

  Summarise the notable changes in the current repo over a time period, from the changelogs and the commits. Use when the user says "codebase news" or "what changed while I was away", or is back from leave.

• constraints

  ★ Constraints: write the project's quality bar to CONSTRAINTS.md, with a command behind each rule, then check each branch against it and catch a lowered bar.

• create-pr

  Push the branch and open a draft PR, or rewrite the body of the open PR. Use when the user asks to open or update a PR, or when the push step of a handoff runs.

• diagnose

  Trace a bug with no obvious cause to a proven root cause, written to a hunt doc. For faults in recent changes where the cause is roughly known, use /fix instead.

• domain-modeling

  Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.

• fix

  Find and fix a broken behaviour from manual QA, or each failed item that a QA report prompt names. Lighter than /diagnose, with no root-cause document.

• happy-path

  Write a manual QA checklist for the current changeset as an HTML report. Use when the user says "happy path" or "test plan", or asks what to test by hand before closing out.

• implement

  Run the next step of a plan handoff with /do-next, test-first, one commit per step. With no handoff, implement a spec or tickets with /tdd at agreed seams. /build runs every step, then tests and reviews in a loop.

• improve-codebase-architecture

  Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

• light-review

  Run the eight essential lanes of the review panel over a PR or branch, into the same report as /engineering:mob-review. Use when the user says "light review" or "quick review", or wants fewer lanes than the full panel.

• mob-review

  Run the full review panel: every in-scope persona lane, plus the project's own rubric, in parallel over a PR or branch diff, into one ranked HTML report. Fixes your own branch, or drafts replies on someone else's. Takes an optional handoff or spec path. Use when the user says "mob review", "run the

• persona-a11y

  Staff accessibility & semantic-HTML reviewer. Reviews changed Vue components for clickable non-interactive elements, icon-only buttons without accessible names, unlabelled inputs, hand-rolled focus/keyboard handling that Reka UI should own, missing landmarks/heading order, and decorative-icon markup

• persona-ac-check

  Staff \"acceptance-criteria coverage\" reviewer. Cross-checks the PR (its code changes AND its description) against the acceptance criteria of a handoff or spec file the caller gives, else of the tracker ticket: flags ACs the PR does not appear to satisfy, and material PR behaviour that no AC covers

• persona-analytics

  Staff product-analytics reviewer for frontends that send events through an analytics wrapper (PostHog today). Reviews a diff for over-tracking (events a backend database already answers, per-keystroke/per-render volume, duplicates of the wrapper's page views or an existing navigation-tracking call,

• persona-api-integrity

  Staff cross-repo API-integrity reviewer for frontend code. For every new or changed frontend API call (a query/mutation factory, an API-client path constant or DTO schema, a legacy API-client method, or a raw fetch), it finds the backend operation that serves it (OpenAPI spec and the controller/hand

• persona-api-spec

  Staff API-contract reviewer. Reviews a diff for drift between the OpenAPI spec and the code that actually serves it: undocumented endpoints, status/enum values the spec omits, response fields whose presence or type no longer matches, stale examples, hand-edited generated files, and spec changes ship

• persona-architecture

  Staff architecture & layering reviewer for frontend apps and shared libraries. Reviews a diff for cross-layer coupling and responsibility placement: business/side-effect logic living in a Vue page or shared shell that belongs in a composable/store/service, shared infrastructure (application frame, l

• persona-async-state

  Staff async-state reviewer for Vue and TypeScript front ends. Reviews a diff for races and stale state around asynchronous work: out-of-order responses overwriting newer input, a request still in flight writing into state that was reset (logout, account or company switch, revert), a reactive route p

• persona-best-practice

  Staff "modern API currency" reviewer. Reviews a diff for outdated Vue/TypeScript/Pinia/Vue-Router patterns where a newer first-class API exists on the installed version (defineModel, reactive props destructure, useTemplateRef, useId, satisfies, no-enum, import type, setup stores, storeToRefs, file-b

• persona-browser-e2e

  Staff e2e reviewer for a standalone browser E2E suite that runs against a real backend, with no MSW (Playwright). Reviews changed specs, page objects, fixtures and API helpers for hard waits, non-retrying assertions, brittle selectors, test isolation, backend-mocking, and page-object/fixture/import-

• persona-bundle-size

  Staff bundle-size / code-splitting reviewer (Vue 3.5, Vue Router 5, Vite 8 on Rolldown). Reviews a diff for bytes that ship to users who never need them: heavy or rarely-rendered components statically imported behind a `v-if` / feature flag / permission / modal / tab instead of `defineAsyncComponent

• persona-charts

  Staff charts reviewer for Vue and TypeScript front ends. Reviews telemetry, usage, and billing charts built on Unovis the way a tester feeds them awkward data: an empty or all-null series, a single point, all zeros, a gap drawn as a drop to 0 or bridged by a straight line, x-axis ticks and tooltips

• persona-code-quality

  Staff code-quality reviewer. Reviews a diff for utility-extraction opportunities, oversized functions, misplaced code, hand-rolled logic that @vueuse/core or es-toolkit already provides, mutation/immutability violations, imperative loops that should be functional, and a new formatter that duplicates

• persona-comments

  Staff commenting reviewer. Reviews a diff for comments that should be removed, what-not-why comments that restate the code, commented-out code, point-in-time comments that will drift stale, redundant JSDoc on internal helpers, doc blocks detached from the declaration they document, noise banners, ov

• persona-correctness

  Staff correctness reviewer for any language. Reviews a diff for logic errors that give a wrong result or a crash: wrong conditions, off-by-one boundaries, crashes on a reachable input, missed cases, the wrong variable or argument order, discarded results and values read too early, callers that the c

• persona-decisions-log

  Staff decision-conformance reviewer. Reads the project's recorded decisions (ADRs in docs/adr/, CONTEXT.md, or a decision log that CLAUDE.md or AGENTS.md names) and checks each changed line against every decision that applies, so a diff does not quietly reverse something the team already settled. Al

• persona-design-system

  Staff design-system/Tailwind reviewer. Reviews a diff for non-semantic color tokens, state/emphasis tokens in the wrong place (a -disabled or -hover token behind the wrong selector, or one that resolves to a no-op), hand-rolled UI that the project's design system already provides, a missing Tailwind

• persona-error-handling

  Staff error-handling reviewer. Reviews a diff for swallowed failures (empty catch, catch-that-only-logs), wrong error routing (console instead of Sentry-for-engineers / toast-for-users), local try/catch that duplicates the TanStack global error handler, errors thrown without context or off the proje

• persona-file-upload

  Staff file-upload reviewer for Vue and TypeScript front ends. Reviews a diff the way a tester attacks an upload: a 0-byte file, a PNG renamed .pdf, SCAN.PDF, a file one byte over the limit, an SVG where a logo should be, the same file picked twice, a PDF dropped beside the zone, three files dropped

• persona-form-behaviour

  Staff form-behaviour reviewer. Reviews a diff the way a tester attacks a form: what Enter, Escape, and the submit button do (@keydown.enter on the design system's input never firing, Enter during IME composition, raw buttons with no type submitting, Enter in a combobox or modal submitting the wrong

• persona-github-actions

  Staff GitHub Actions reviewer (frontend and backend repos alike). Reviews changed workflow and composite-action files for CI/CD security and best-practice defects: script injection via ${{ }} interpolation of untrusted event data into run blocks, over-broad or missing permissions, mutable action ref

• persona-i18n

  Staff i18n reviewer for an app that uses vue-i18n. Reviews a diff for hardcoded user-facing strings, dynamic translation keys, product/brand names not injected as linked refs, duplicate keys for text that already exists, new namespaces, wrong Lokalise leaf format, HTML in translation strings, and mi

• persona-integration-test

  Staff integration-test reviewer for frontend apps (Playwright + MSW). Reviews changed e2e tests for hard waits, non-retrying assertions, brittle selectors, MSW handler-reset/isolation problems, weakened safety nets, test interdependence, and page-object/fixture bypass. Use when the user asks for an

• persona-naming

  Staff naming & file-placement reviewer. Reviews a diff for file/folder naming conventions, files living in the wrong directory, export naming, and identifier naming (meaningful names, unit-suffixed numerics, consistent vocabulary for one concept, and boolean/composable naming). Use when the user ask

• persona-pr-size

  Staff \"PR size & splittability\" reviewer. Judges whether a PR/branch diff is too big to review well, whether it decomposes into independent or sequential slices, and, when it does, proposes the concrete stack layout (branch names bottom→top, what each slice holds, the dependency order) to build wi

• persona-production-protector

  Staff \"production protector\" reviewer. One job: keep the change's production blast radius as small as possible. Checks that work which can sit behind a feature flag does (and that the gate is complete, fails closed, and is a known flag), and, beyond flags, that the diff doesn't change live behavio

• persona-resilience

  Staff resilience reviewer for Vue and TypeScript front ends. Reviews a diff for what happens to a task when the session, the network, another tab, or a deploy changes under it: a 401 or session expiry that throws away a half-filled form, parallel 401s each running their own logout or refresh, 4xx er

• persona-security

  Staff security reviewer for Vue/TypeScript frontends. Reviews a diff for security defects: reverse-tabnabbing (target=_blank / window.open without noopener), wildcard postMessage and unvalidated message origins, v-html / unsanitized HTML injection (XSS), dev/feature-flag overrides readable in produc

• persona-shell

  Staff shell-script reviewer. Reviews changed .sh/.bash files and multi-line workflow `run:` blocks for correctness defects that make a script fail open or fail silently: exit statuses swallowed by process substitution, pipelines, `local`, or `if`; missing `set -euo pipefail`; SIGPIPE traps under pip

• persona-skills

  Staff reviewer for Claude Code skills, plugins and agent instructions. Reviews changed SKILL.md files, skill references, agent/command definitions, CLAUDE.md/AGENTS.md and plugin/marketplace manifests for defects that make an agent following them fail or misbehave: dangling references to skills, too

• persona-tanstack

  Staff TanStack Query / data-fetching reviewer. Reviews a diff for query-key correctness, cache invalidation bugs, staleTime/gcTime misuse, reactivity of query args, mutation invalidation, optimistic-update safety, and API-client usage in queryFns. Use when the user asks for a TanStack/vue-query/data

• persona-test-design

  Staff \"think like a tester\" reviewer for frontend and backend PRs. Reviews what SHOULD be tested rather than how tests are written: rates the risk of each changed behaviour (likelihood × impact), derives the test conditions a tester would demand (equivalence partitions, boundary values, decision t

• persona-test-quality

  Staff test-quality reviewer for test code in any language. Reviews changed tests for coupling to the implementation (private methods, internal state, checks through a side channel), mocks of the project's own code, expected values that restate the implementation, tests that pass when the feature bre

• persona-text-input

  Staff text-input reviewer. Reviews a diff for what happens to the characters a user types or pastes: whitespace-only values passing required, trim applied to passwords, over-strict name/email/key validators that reject real input, comparisons that ignore case/whitespace/Unicode normalisation, autoca

• persona-typing

  Staff type-readability reviewer. Reviews a diff for types a human can't read at a glance: casts (`as`, `as unknown as`, `!`, `any`) that hide what a value actually is, derived-type gymnastics (`ReturnType<>`, `Parameters<>`, `(typeof X)[number]` chains), `Omit<>`/`Partial<>` pyramids that describe a

• persona-url-state

  Staff URL-state and navigation reviewer. Reviews a diff for what refresh, a shared link, Back, and Forward do to a screen: filters, search, and tabs held in refs instead of the URL, query params read once so Back changes the URL but not the screen, router.push per keystroke or on a redirect/cancel (

• persona-value-display

  Staff value-display reviewer. Reviews a diff for what a value looks like on screen when it's missing, zero, huge, or awkward: 'undefined Mbps', NaN, 'Invalid Date', and [object Object] rendered as text; a null date shown as 1970 or 2000 and an unknown price shown as $0; placeholders that disagree an

• persona-vitest

  Staff Vitest unit-test reviewer. Reviews changed unit tests for flaky-test risks, accessible-selector usage over test ids, proper mock-data builders, mock leakage, one-assertion-per-waitFor, describe.each/helper dedup, test-runner API currency against the installed vitest version, and whether each t

• persona-vue-reactivity

  Staff Vue reactivity reviewer. Reviews a diff for reactivity correctness and cost: static values needlessly wrapped in ref/computed, watchers that should be computed, lost reactivity from destructuring, unstable props causing needless re-renders, missing/unstable v-for keys, ref-vs-reactive/shallowR

• pick-next-task

  Rank the open work in your team's backlog, then assign the tickets that you pick. Use when the user says "pick next task" or asks what to work on next.

• pickup

  Pick up an open PR: claim its worktree, sync it, and assess its open review threads into a report. Use when the user asks to pick up a PR from its number, URL, branch, or ticket key.

• plan

  ★ Plan: turn a ticket or a spec into a claimed worktree and an approved plan handoff in vertical slices, through /start-ticket.

• post-review-comments

  Post the rows ticked in a review report to the PR as one review. Use when the user pastes the prompt that a review report builds, or asks to post the ticked findings.

• pr-news

  Summarise what happened on your open PRs and your review requests over a time period, one table per repo. Use when the user says "pr news" or "catch me up on my reviews", or is back from leave.

• prototype

  Build a throwaway prototype to answer a design question about logic or UI.

• qa-screenshots

  Run a happy-path QA checklist in your own browser, save a screenshot for each item, show them in the QA report, and append a Manual QA section to the PR.

• refine-ticket

  Refine a ticket before work starts into lettered ACs and a plain summary, written back after one yes. Use when the user asks to refine a ticket, or when another skill finds a ticket with no user story, no ACs, or vague scope words.

• resolve-conflicts

  Resolve git merge/rebase conflicts in a file using its history and the in-progress operation as context. Use when the user says "resolve conflicts", "fix the merge", "resolve this conflict", or hands you a conflicted file.

• resolve-feedback

  Assess review feedback from AI or human reviewers, pasted or from the open threads of a PR. Write a verdict, a recommendation, and a draft reply for each item to a review report, whose ticks build the /engineering:apply-review prompt. Applies no fix.

• review-pr

  Review an open PR in its own worktree: claim the worktree, sync it, and run /engineering:mob- review there. Use when the user asks to review a PR from its number, URL, branch, or ticket key.

• review

  ★ Review: review a branch or a PR. The current branch gets /light-review, or /mob-review when you say full. A PR number, URL, or ticket key goes to /review-pr.

• sad-path

  Write a manual QA checklist that tries to break the current changeset, from the test data of the edge-case personas. Use when the user says "sad path", asks which edge cases to try by hand, or wants to break a form.

• scope-ticket

  Scope a ticket before work starts, with one verdict (SCOPED, SPLIT, or CLARIFY) and a points estimate. Use when the user asks to scope a ticket or if it fits one PR, or when another skill needs a scope check.

• self-improve

  Mine a repo's fix commits, tracker bugs, review threads, failed CI runs, and the calibration backlog, put every proposal to you in one table, apply the approved persona edits through update-personas, and write a report and the repo's defect hotspots.

• ship

  ★ Ship: check that the branch is ready (plan steps, tests, review, constraints), give a go or no-go, then open or update the PR through /create-pr.

• spec

  ★ Spec: define what to build. A ticket key goes to /refine-ticket. A raw idea gets an interview to an explicit yes, then a written spec that /plan reads.

• split-ticket

  Split a ticket that is too big for one PR into sub-tasks in waves, with a stub handoff for each. Use when the user asks to split a ticket, or when a scope check gives SPLIT and the user picks the split path.

• squash-bugs

  Fix a batch of QA bug tickets in parallel, one worktree and one subagent per ticket, then report every fix and every open decision.

• start-ticket

  Start work on a ticket: refine, scope, or split it as needed, claim its worktree, and write the plan handoff. Use when the user asks to start or pick up a ticket or a sub-task.

• sync-docs

  Update every doc, README, type, and config that the current code changes put out of sync.

• tdd

  Test-driven development with the red → green loop. Use when the user wants to build a feature or

• update-personas

  Refresh the persona-* reviewers: check persona-best-practice against the installed versions, and lint the other personas. Use when the user says "update personas" or the personas are stale, or after a big dependency bump.

• webperf

  ★ Webperf: measure a web page's performance first, find the one bottleneck, fix it, then measure again and keep or revert the fix.

Similar plugins