What's in the toolkit

Backend architecture selection and cross-language fix playbooks. Use when starting a backend, choosing stack/database/architecture, designing APIs, or diagnosing and fixing backend bugs and performance issues in any language (Node, Python, Go, Java, PHP, Rust, C#, Ruby).

Context-overload guard. Use when the session context is getting heavy — long conversation, repeated file re-reads, degraded recall of earlier decisions, or the user says "context", "session is slow", "you're forgetting things", or asks to hand off to a new session. Produces a handoff brief and tells the user to start fresh.

World-class animation and motion design. Use when building or reviewing any animation — page transitions, micro-interactions, hover states, loaders, scroll effects, 3D (Three.js), physics, or when the user says "animate", "smooth", "polished", "delightful", "like Apple/Stripe/Linear".

Turns raw ideas into build-ready PRDs. Use when the user describes a feature or product idea, says "I want to build...", asks for requirements, user stories, acceptance criteria, or scoping — the missing front end of the dev lifecycle before any code is written.

Engineer-mind learning loop — verify current tech before using it, detect syntax/API drift, and record learnings. Use when working with a framework version newer than training data, when APIs behave unexpectedly, when the user mentions "latest", "new version", "deprecated", or asks what's new in a technology.

Testing strategy and execution across stacks. Use when writing tests, fixing failing tests, setting up test infrastructure, handling flaky tests, or when the user mentions coverage, unit/integration/e2e testing, TDD, or QA.

Patch- and ticket-based development workflow. Use when handling a bug ticket, issue, hotfix, small fix, or patch — anything referencing Jira/Linear/GitHub Issues IDs (e.g. "PROJ-123", "fix

Token-efficiency rules for every task. Use at the start of any coding session, when the user mentions tokens, cost, context limits, or when work involves large files/codebases where careless reading would burn context.

World-class UI/UX design system knowledge. Use for any interface work — building screens/components, choosing color palettes, layouts, typography, spacing, accessibility, or when the user says "make it look good/professional/modern", mentions design, UX, redesign, landing page, dashboard, or mobile app UI.

Frontend & mobile performance optimization. Use when pages load slowly, bundles are big, animations jank, Lighthouse/Core Web Vitals scores are low, or the user mentions performance, speed, FPS, bundle size, or memory on the client side.

Polyglot backend specialist. Use PROACTIVELY for API design, database work, performance issues, and backend bugs in any language (Node, Python, Go, Java, PHP, Rust, C#).

Ticket-driven minimal-diff patcher. Use for bug tickets, hotfixes, patches, and small fixes — especially when a ticket ID (PROJ-123,

Use PROACTIVELY as a review gate on any non-trivial diff before it is declared done or merged. Reviews for correctness, security, performance, and design-system compliance.

Documentation specialist. Use for READMEs, API docs, onboarding guides, architecture decision records (ADRs), and code comments — when docs are missing, stale, or requested.

Handles releases, versioning, changelogs, and deploy readiness. Use for "ship it", "release", "deploy", version bumps, tagging, and hotfix releases.

Deep security review beyond the code-reviewer's pass. Use before releases, after auth/payment changes, when adding dependencies, or on request ("is this secure?"). Read-only — reports, never edits.

Researches current tech — new framework versions, syntax changes, deprecations, security advisories. Use when APIs behave unexpectedly, before major upgrades, or when the user asks "what's new in X" or "should we adopt Y".

Use PROACTIVELY to run tests and fix test failures after code changes. Runs the right framework automatically, fixes failing tests without changing product logic.

Frontend + design specialist. Use PROACTIVELY for building or restyling screens, components, landing pages, dashboards, and mobile UIs, and for design reviews of visual work.

PostToolUse hook wiring: Prettier/black on every edit — zero tokens on formatting (see settings-hooks.json).

PreToolUse guard: blocks rm -rf /, force-push, DROP TABLE and other destructive commands before execution.

Stop hook: appends a session entry (date, approx tokens) to activity/changelog.jsonl.

Resolve a ticket end-to-end (reproduce → fix → test → PR → close loop)

Build a feature end-to-end from an idea or PRD (plan → build → test → review)

Pre-release pipeline — test, review, version, changelog, tag

Daily standup summary — what shipped, what's in flight, what's blocked

Audit a screen/component against the ui-ux-master + motion-animation checklists