Teach your agent new tricks

Catches missing ARIA labels, broken keyboard navigation, trapped focus, insufficient contrast, and missing screen reader announcements. Thinks like a user who can't see the screen or use a mouse.

Catches unsafe Rust patterns, error handling gaps, resource leaks, concurrency issues, and IPC protocol mistakes. Reviews Tauri commands, Rust crates, sidecar processes, and file system operations.

Catches visual inconsistencies, design system violations, and craft issues. Reviews UI changes, component styling, visual states, dark mode, colors, typography, spacing, and animations.

Catches usability issues specific to developer tools. Reviews CLI commands, error messages, configuration, onboarding, keyboard shortcuts, tool output, and API surfaces.

Catches React anti-patterns, state management issues, type safety gaps, unnecessary re-renders, and bundle bloat. Reviews React components, hooks, Zustand stores, TypeScript types, and CSS/Tailwind.

Systems performance — memory over 8-hour sessions, IPC latency, streaming throughput, resource lifecycle. Reviews event listeners, state subscriptions, process spawning, and long-lived session stability.

Catches failures that only appear in production — environment differences, missing error recovery, deployment gaps, and data integrity issues. Reviews persistence, process lifecycle, and rollback safety.

Catches the slow erosion of codebase health. Reviews barrel exports, file placement, naming conventions, import ordering, style reuse, code deduplication, and lint compliance.

Catches coupling violations, abstraction mismatches, circular dependencies, and architectural drift. Reviews module boundaries, dependency graphs, layer crossings, and system reorganizations.

Catches usability issues, broken mental models, missing feedback, and interaction patterns that look fine in code but confuse real users. Reviews user journeys, flows, navigation, onboarding, and error states.

Teaches the thinking process, not just rules. Study the codebase first, plan properly, implement with discipline. Asks the questions a senior engineer asks before every decision — follow existing patterns, find root causes, never hardcode or patch around problems.

Two-phase workflow: analysis then generation. Captures HTML elements from DevTools, locates source components, discovers edge cases, and produces production-grade Vitest + React Testing Library tests.

Gates plans before they enter the audit pipeline. Hunts for buried scope reductions, scope reduction by omission, vague confidence language, and approach issues like building new when extending existing systems would work. Binary PASS/FAIL with severity-rated fix list.

The nuclear option. Runs all 10 specialist audit lenses sequentially — design, UX, frontend, backend, structural, repo, production, DX, performance, and accessibility — writing everything into one comprehensive document. Works in plan mode or repo mode.

Creates in-app stress tests that drive the real application pipeline from browser DevTools. No mocks — real state, real APIs, real UI. Works for any web app with observable state.

Clarifies problems, bugs, and feature requests through a reflect-back interview before entering plan mode. Ensures you and the AI agent have the exact same understanding of the problem before any plan is written.

Creates in-app stress tests that run from browser DevTools console. Drives the real message pipeline with no mocks — real Tauri events, real Zustand stores, real Rust backend.

Writes tests that fail when implementation breaks. No placeholders, no expect(true).toBe(true). Supports TypeScript/React with Vitest and Rust with cargo test.

Comprehensive performance optimization guide for React and Next.js applications. Contains 45 rules across 8 categories, prioritized by impact to guide automated refactoring and code generation.

Reviews UI code against Web Interface Guidelines. Checks accessibility, design patterns, and best practices, then outputs findings in a terse file:line format.