imaimai-front-templete

The pols.dev anti-slop design law — a comprehensive catalog of generic "AI slop" UI patterns to avoid (glowy pill buttons, gradient heroes, kitchen-sink cards, default font pairings, botched glass/shadows/glows, slop layouts) and what premium craft looks like instead. Use EVERY time you design, build, restyle, or review any user interface, landing page, or visual component — read it before starting, and re-check the output against it before calling the work done.

Unified animation and motion skill — Apple-style fluid interface design, Emil Kowalski's craft standards, and an animation vocabulary glossary. Use when designing, building, or reviewing any animation, transition, gesture, or motion effect. Covers springs, easing, gestures, interruptibility, performance, accessibility, and a reverse-lookup glossary for naming effects.

Use when finishing a feature, fixing a bug, before committing React code, or when the user types `/doctor`, asks to scan, triage, or clean up React diagnostics. Covers lint, accessibility, bundle size, architecture. Includes a regression check and a full local-triage workflow that fetches the canonical playbook.

On-demand repo-wide audit with the best available model — hunts what deterministic gates cannot (architecture drift vs ADRs, security posture, dependency strategy, doc staleness/DX) and routes outcomes into existing rails (ADR/AGENTS.md via aegis-share, or a plan doc in docs/superpowers/specs/). Use when the user asks for a repo audit, health check, or "what should we improve"; not scheduled, not CI-run.

Unified pre-commit review of the uncommitted diff — a finder agent hunts across all lenses (bugs + AGENTS.md rules), then a separate verifier agent adversarially refutes each finding, and its completion stamps the commit gate (ADR-0009/0011/0015). Run before every commit, or whenever the uncommitted diff needs a full review. Pass "high" for a deeper multi-lens verify pass.

Explicit entry point for ticket-granularity work (implement a component, fix a non-trivial bug, refactor a module, add a feature). The user invokes /start-workflow at the start of such tasks. The skill enforces an orchestration sequence — clarify → aegis → plan → optionally ADR → implement (parent by default, delegate by context impact) → fresh-context review → commit/PR — so the work follows the project's rules-driven standards. Skip for one-line fixes, config tweaks, docs-only changes, and trivial typo fixes — handle those directly.

Agent-based model checking of a state-machine spec — a hunter agent formalizes + hunts counterexamples, then a separate checker agent replays each against the machine (flat pipeline, ADR-0010/0015). Design step of start-workflow for interaction-complex features (wizards, auth/session flows, async guards, permission branching). Pass the spec path as the argument.

Use when forking this template for a project that does not need a database, auth, or file storage — removes Cloudflare D1 / R2 / Better Auth / Drizzle ORM and the Cloudflare Workers deployment wiring.

Pre-launch checklist for web services. Comprehensively audits security, SEO, OGP, performance, accessibility, and more, then generates a report.

Run Lighthouse audits (a11y, SEO, best practices) across all pages for desktop and mobile, then record results in docs/lighthouse/audit/.

Run performance traces (Core Web Vitals — LCP, CLS, INP) across all pages, analyze insights, record results in docs/lighthouse/performance/, then fix issues found.

agent 向けテキスト指示（skill / slash command / task プロンプト / CLAUDE.md 節 / コード生成プロンプト）を、バイアスを排した実行者に動かしてもらい、両面（実行者の自己申告 + 指示側メトリクス）で評価して反復改善する手法。改善が頭打ちになるまで回す。プロンプトや skill を新規作成・大幅改訂した直後、またはエージェントの挙動が期待通りにならない原因を指示側の曖昧さに求めたいときに使う。