engineering

Ultimate Skills

One opinionated, deduplicated marketplace of the best Claude Code skills — distilled from five community skill repos (109 skills → 39 keepers) and regrouped into four discipline plugins: engineering · product · design · research.

This is a curation, not original work. Every skill keeps its original author and license — see NOTICE.md and PROVENANCE.csv. It was assembled by a small team of specialist reviewers (engineer, product/marketing, UX, repo expert) with a devil's-advocate pass to cut redundancy.

What's inside

Plugin	Skills	License	Focus
engineering	16	MIT	Spec → plan → build → test → debug → review → harden → ship
product	13	MIT	Discovery, strategy, prioritization, PRDs, pricing, metrics
design	6	MIT	Anti-slop frontend taste, image-to-code, redesign, UI engineering
research	4	CC-BY-NC-4.0	Deep research, academic writing, peer review

⚠️ Mixed licensing. engineering, product, and design are MIT (commercial use OK). research is CC-BY-NC-4.0 — non-commercial only, isolated in its own plugin so it can be dropped cleanly. See NOTICE.md.

Install

# Add this marketplace (from GitHub)
/plugin marketplace add Liohtml/ultimate-skills

# …or from a local clone
/plugin marketplace add /path/to/ultimate-skills

# Then install the plugins you want
/plugin install engineering@ultimate-skills
/plugin install product@ultimate-skills
/plugin install design@ultimate-skills
/plugin install research@ultimate-skills   # CC-BY-NC — non-commercial

Each skill activates automatically when its trigger matches your task; you can also name it explicitly (e.g. "use the strategy-red-team skill on this PRD").

Skill catalog

engineering (MIT)

Skill	What it does
spec-driven-development	Write a spec before coding when requirements are unclear or work is significant.
planning-and-task-breakdown	Break a spec into ordered, independently verifiable tasks.
incremental-implementation	Ship multi-file changes as small, safe vertical slices.
test-driven-development	Drive logic and bug-fixes with tests; prove code works.
debugging-and-error-recovery	Systematic root-cause debugging when things break.
code-review-and-quality	Multi-axis review with severity labels before merge.
code-simplification	Refactor for clarity without changing behavior.
security-and-hardening	Harden input handling, auth, storage, and integrations (incl. LLM).
performance-optimization	Measure-first optimization, budgets, Core Web Vitals.
api-and-interface-design	Stable, contract-first API and module-boundary design.
git-workflow-and-versioning	Commits, branching, conflicts, bisect-driven debugging.
ci-cd-and-automation	Quality-gated build/deploy pipelines and rollout/rollback.
doubt-driven-development	In-flight fresh-context adversarial review of key decisions.
documentation-and-adrs	ADRs, READMEs, and changelogs that record context.
browser-testing-with-devtools	Verify behavior in a real browser via Chrome DevTools MCP.
karpathy-guidelines	Lightweight behavioral rules to avoid common LLM-coding mistakes.

product (MIT)

Skill	What it does
strategy-red-team	Steelman then attack a PRD/roadmap's load-bearing assumptions.
opportunity-solution-tree	Torres-style discovery: outcome → opportunities → solutions → experiments.
prioritization-frameworks	9 frameworks (RICE, ICE, Kano, MoSCoW…) with formulas and when-to-use.
pre-mortem	Imagine the launch failed; triage Tigers / Paper Tigers / Elephants.
create-prd	8-section Product Requirements Document.
pricing-strategy	Pricing models, value metric, Van Westendorp, experiments.
ab-test-analysis	Significance, sample size, SRM, guardrails → ship/extend/stop.
north-star-metric	Define the NSM + input-metric constellation by business game.
value-proposition	6-part JTBD value-proposition design.
competitor-analysis	Research-driven competitive landscape and differentiation.
interview-script	Mom-Test / JTBD customer-discovery interview script.
user-personas	Research-grounded personas with JTBD, pains, gains.
intended-vs-implemented	Audit the gap between intended behavior and shipped code.