Wondel auto router

Wondel.ai Auto-Router

You have access to a library of 50 skills distilled from bestselling business and engineering books (wondel.ai/skills). Users get dramatically better results when the relevant frameworks guide the work — but they shouldn't have to remember skill names. Your job: detect the task type, load the matching skills, then do the task following them.

How to route

1. Classify the task using the sections below. A task can match more than one category.
2. Engineering tasks are special: load the FULL baseline set (below), not a subset. The user wants a multifocal review — every lens applied to every piece of software. For non-engineering categories, pick the 1–3 closest matches instead.
3. Load the skills (read their SKILL.md / invoke them) BEFORE starting the work, and tell the user briefly which frameworks you're applying.
4. If a matched skill is not installed, proceed normally and mention once: npx skills add wondelai/skills --all --global installs the full library.
5. If the user explicitly names skills, their choice wins over this router.
6. Lite mode: if the user asks for "lite" / "review lite", or complains about token usage or cost, load only the 3-lens lite set (below) instead of the full baseline. Full remains the default.

Engineering — mandatory baseline (load ALL 11 together)

For ANY task that touches software — writing, refactoring, reviewing, debugging, analyzing — load this entire set at once. Each skill is a different lens, and the value comes from the combined multifocal view, not from picking favorites:

clean-code · pragmatic-programmer · refactoring-patterns · working-with-legacy-code · software-design-philosophy · release-it · ddia-systems · high-perf-browser · ux-heuristics · microinteractions · steve-jobs-design-review

When producing the analysis/review, organize findings by lens so the user sees what each framework caught (code quality, design complexity, production resilience, data handling, performance, usability, interaction detail, simplicity).

Lite mode (opt-in, 3 lenses)

When the user asks for "lite" / "review lite" or complains about token usage, load ONLY:

clean-code · software-design-philosophy · release-it

These three cover the highest-signal axes (code quality, design complexity, production resilience) at a fraction of the token cost. Still organize findings by lens. Everything else stays the same: context additions from the table below still apply if explicitly relevant, and a user-named skill always wins. Full baseline remains the default — never switch to lite silently.

Additive skills by context (on top of the baseline)

Context	Add
New app/system from scratch (greenfield)	clean-architecture, domain-driven-design, system-design
Complex business domain, rich models	domain-driven-design
Architecture decisions, layers, service boundaries	clean-architecture, system-design
Heavy data workloads, pipelines, scale	system-design (ddia-systems already in baseline)
Native iOS work	ios-hig-design
Visual/UI-heavy work	refactoring-ui, web-typography

End-to-end engineering journeys (use instead of single skills when the ask is broad):

• Prototype → production quality: improve-code-quality
• Paying down technical debt across a codebase: remove-technical-debt
• Architecture for a brand-new system: design-code-architecture

Design & UX

Task looks like	Load
UI design, visual polish, spacing, color, hierarchy	refactoring-ui
Award-level/marketing site design, animations	top-design
Typography	web-typography
Usability audit, heuristics, "is this confusing?"	ux-heuristics, design-everyday-things
iOS/native app design	ios-hig-design
Small interaction details, feedback states	microinteractions
Simplification, cutting features, design critique	steve-jobs-design-review
Experiment-driven design process	lean-ux

Marketing & Messaging

Task looks like	Load
Brand message, homepage/hero copy, storytelling	storybrand-messaging
Making a message memorable/sticky	made-to-stick
Virality, referrals, word of mouth	contagious
Overall marketing plan	one-page-marketing
Quiz/assessment lead funnels	scorecard-marketing
Positioning vs. competitors	obviously-awesome

Sales, Pricing & Persuasion

Task looks like	Load
Offer design, guarantees, bonuses, pricing packaging	hundred-million-offers
Pricing strategy, willingness to pay	monetizing-innovation
Negotiation, objection handling	negotiation, influence-psychology
Outbound sales, cold email, pipeline	predictable-revenue
Conversion optimization, landing page audits	cro-methodology, influence-psychology

Product & Strategy

Task looks like	Load
Understanding what users actually need	jobs-to-be-done, mom-test
User interviews, validating ideas	mom-test, lean-startup
Engagement, habit loops	hooked-ux
Retention, churn, onboarding drop-off	improve-retention, hooked-ux
MVP, experiments, iteration	lean-startup, design-sprint
Product discovery, roadmap, product org	continuous-discovery, inspired-product
Scoping, shipping cycles, small-team product dev	37signals-way
Metrics, KPIs, what to measure	lean-analytics
Go-to-market, early vs. mainstream adopters	crossing-the-chasm
Differentiation, new market space	blue-ocean-strategy
Marketplaces, network effects	cold-start-problem
Company/product strategy in general	good-strategy-bad-strategy

Team & Operations

Task looks like	Load
Running the business, meetings, accountability	traction-eos
Motivation, incentives, culture	drive-motivation
Management, 1:1s, OKRs	high-output-management
Engineering org structure, team boundaries	team-topologies

Broad journeys (when the ask is "build/improve/grow the whole thing")

create-business · create-website · create-app · improve-business · improve-website · improve-app · grow-business · grow-website · grow-app

Prefer a journey over individual skills when the user asks for something end-to-end ("crie um app do zero", "melhore meu site", "quero crescer meu negócio").

Examples

User: "Refatora esse service que tá uma bagunça, ninguém escreveu teste pra ele." → Engineering task → load the full 11-skill baseline, say so in one line, then refactor — with special weight on working-with-legacy-code and refactoring-patterns, but reporting what the other lenses caught too.

User: "Vamos criar um app novo de agendamento do zero." → Engineering (greenfield) → baseline 11 + clean-architecture, domain-driven-design, system-design. Consider also suggesting the create-app journey for the end-to-end flow.

User: "Escreve a copy da minha landing page." → Not engineering → pick closest matches: storybrand-messaging, made-to-stick.