gbasin

Decomposed reasoning with explicit confidence scoring

Adversarially stress-test a technical plan by verifying claims against real docs, running POC code, and updating the plan before you build.

Hand-computes a system's execution with explicit state at each step. Use when debugging races, state machines, async bugs, regressions; stress-testing a fix; scoping features on existing state; or approaching unfamiliar code. Especially useful for bugs involving client/server sequencing, background refreshes, recovery paths, queues, or multiple actors evolving state together. Also triggers on "why is this broken", "walk through the flow", "this used to work", "what happens when", or "stress test this plan".

Detect and diagnose visual jank in React UIs. Covers two modes -- static analysis of code patterns that cause flicker, layout shift, and flash of intermediate state (for pre-merge review or preventive sweeps), and runtime investigation of specific visual symptoms (flicker, blink, jump, stutter, remount churn, scroll reset, post-paint correction). Use when the user mentions jank, something feeling off, flashing, blinking, layout shifting, full repaints, or wants a render stability audit.

Catches state bloat, grab-bag models, and mutation ambiguity from AI coding agents. Use when reviewing state types, boolean flags, optional-field models, or mutable data patterns.

Convert Figma designs to pixel-perfect React components with Tailwind CSS.

Delegate implementation work to parallel headless agent-CLI subagents (codex by default; omp with a cheap model like Gemini Flash as an alternative) in persistent git worktrees, with the current agent as orchestrator (plan, review, merge, final QA). Use when the user asks to delegate to codex or omp, fan out work to subagents, or parallelize implementation across cheaper agents. Includes durable runner supervision and the recipe for subagents doing their own visual QA via dev-browser.

Evidence-grounded first-run UX/UI audit of a running web app, through pluggable design lenses (Steve Jobs, Nielsen heuristics, Dieter Rams, WCAG). Drives the real UI with dev-browser, fans out parallel analysts, verifies findings against source, renders Today/Refine/Bold solution mockups, and ships an interactive, hosted HTML walkthrough that collects your direction picks back. Use when the user wants a deep UX/design critique, first-run/onboarding review, heuristic evaluation, "audit the app through a Steve Jobs lens", or a comparative redesign-direction report.

Apply the 2007 iPhone launch's rhetorical product-pitch architecture to create persuasive decks, websites, launch narratives, sales materials, demo scripts, or stories for other products. Use when a product needs sharper category framing, status-quo contrast, mechanism reveal, proof-by-demo, value anchoring, or a more memorable close. Avoid cosmetic Apple imitation; adapt the structure to the target product, audience, proof, and medium.