Use this skill for complex, multi-step, ambiguous, research-heavy, implementation-heavy, cross-domain, high-volume, or high-impact tasks where autonomous decomposition, specialist delegation, parallel investigation, or independent verification would materially improve quality or speed. Proactively manage the full manager-worker loop without requiring the user to request subagents, reviewers, auditors, retries, or synthesis. Keep simple one-step work in the parent agent when delegation overhead would exceed its value.
Use this skill when a task is quality-sensitive, complex, reusable, high-impact, or explicitly asks to refine, improve, critique, debug, verify, optimize, red-team, polish, or make an output production-ready. Apply an adaptive define-build-evaluate-revise-verify loop to writing, code, research, analysis, plans, prompts, designs, decisions, and files. Do not activate for trivial one-step answers, casual conversation, or simple transformations unless the user explicitly requests iteration.
Transform long, messy, dictated brainstorming sessions into evidence-grounded, de-duplicated, challenged, and executable plans through an interactive convergence workflow. Use when a user provides a brain dump, stream-of-consciousness notes, scattered ideas, competing options, assumptions, or an abstract objective and wants factual research, synthesis, prioritization, decision support, and concrete next actions while preserving the original essence.
Build grounded, artifact-aware, verification-first execution plans for complex user goals. Use when the user asks to build, launch, implement, migrate, research, architect, roadmap, strategize, decompose, or create a to-do list for any substantial project. Produces pragmatic plans that balance state modeling, hierarchical decomposition, assumptions, risks, dependencies, speed, and measurable execution gates without generic filler.
Orchestrate a premium polish pass across code, design, image-generation concepting, animation, smoothness, performance, copy, architecture, and opt-in accessibility. Use when the user asks for sexyness, sexyfication, sexification, make it sexy, make it premium, remove AI slop, use imagegen to re-imagine a design, make code/design/UI/UX feel elite, smooth, beautiful, high-quality, delightful, optimized, or competition-grade. Loads nested sexyness sub-skills from subskills/*/SKILL.md as needed and applies QA gates before declaring work done. Accessibility is only selected when explicitly requested or when the active task is specifically an accessibility audit/fix.
Create polished, non-generic visual design with strong hierarchy, spacing, typography, color, alignment, density, design-system fit, and optional imagegen-assisted visual direction. Use when improving UI, pages, dashboards, components, layouts, visual rhythm, brand feel, or anything that should stop looking like AI-generated slop.
Use image generation as a premium design accelerator for UI re-imagining, visual direction, mockups, hero/product assets, game visuals, and bitmap concept references. Use when the user asks to use imagegen, generate visual variants, re-imagine a design, or when generated imagery can materially improve a visual implementation before code is changed.
Make premium work inclusive and robust: keyboard access, screen reader semantics, focus management, contrast, target size, reduced motion, error messaging, and resilient content. Use only when the user explicitly asks for accessibility, a11y, WCAG, screen reader support, keyboard support, contrast, reduced motion, inclusive design, or an accessibility audit/fix. Do not trigger from general sexyness, premium UI, design polish, or animation requests.