Art & Design

Turn a user-provided video into an Apple-style, scroll-driven website with smooth canvas video scrubbing, optimized WebP frame sequences, progressive loading, and responsive motion. Use for cinematic product reveals, landing pages, scrollytelling, image-sequence animation, and scroll-controlled video; also optimize generated frame assets.

Develop software through a separate domain architect and implementation agent, ADRs, and a graph of evidence. Use for ADR-driven development, architecture decisions requiring explicit user approval, implementation, and independent verification. Supports business and technical domains in Codex and Claude Code.

智能戒指 / 健康穿戴品类「卖点文案 Guideline + 速评卡 + 改稿法则」。用于撰写、诊断、打分、改写标题 / 小标题 / 一句话卖点，以及审稿改稿，适用详情页、视频字幕、Meta/TikTok 广告、EDM 主题行、Google 搜索标题。触发词：「智能戒指文案」「戒指标题」「卖点文案」「标题速评」「文案打分」「标题诊断」「改稿」「审稿」「文案校正」「工程师语言」「翻译腔」「smart ring copywriting」「headline review」「copy scoring」「revision」。纯品牌口号创作与非健康穿戴品类不适用（规则可借鉴，品类词表不适用）。

Entity resolution and record matching — find which records across one or more datasets refer to the same real person, company, place, or thing, then merge them into one trusted "golden record" with full source history. Use this skill whenever the user wants to deduplicate a list, merge customer/patient/contact/vendor files, match records without a shared ID, link addresses, clean a CRM, build a master data (MDM) pipeline, score fuzzy name matches, choose blocking keys, set match thresholds, or a

Gives every change to AI-built work its own version number (v1.2.3), keeps every version until the user deletes it, and lets the user compare versions or go back without losing work. Use whenever you create or edit files for a user over several turns (designs, web pages, apps, documents, code), even if they never mention versions, and whenever the user asks to save a version, see history, compare, undo, restore, roll back, or go back to an earlier or "the previous" version.

Cost-optimized task segregation for building products and features. Captures the main task, saves it to a resumable ledger, splits it into atomic subtasks, scores each subtask's difficulty, routes it to the cheapest model + effort that can do it well (inline shell / haiku / sonnet / opus), batches subtasks that share the same model + effort into ONE subagent thread, runs independent batches in parallel waves, verifies with deterministic checks and escalates only failures one tier up. TRIGGER whe

Build web UI in a quiet industrial style distilled from game-menu conventions (paper and ink, one accent for "current" and "primary", selection that reads at a glance, pill secondary buttons, `//` path titles, faint closed contour lines). Use when asked for a page, portfolio, dashboard or component in this style, or in an "industrial", "terminal-like", "Endfield-inspired" or 终末地风格 look.

Release a new simctlhouse version end to end - bump the version, agree on release notes with the user, build the signed and notarized tarball with scripts/release.sh, tag, create the GitHub release, and publish the formula to the Homebrew tap. Use when asked to release, publish, ship or cut a simctlhouse version, or to update the Homebrew formula.

Improve a product's design in rounds until it is ready for real people: every user journey scores 9 or better from independent AI reviewers made by different companies, who agree within a point, with no high-severity fix open, held for two rounds. Uses at least two reviewers (for example Claude and Codex). Each round captures the same screens the same way, each reviewer drives the real product in a real browser in fresh context and scores every journey 0 to 10 with an "if fixed" score and specif

Security and engineering red lines for designing, writing, or reviewing any website, backend, or API: hardened auth, injection prevention, business-logic and race-condition defense, high-concurrency database design, async-first architecture, API integrity and anti-tampering, secrets and operations, AI/LLM security, mobile and mini-program security, CORS and security headers, and scenario-aware 404 and indexing rules. 系统与网站开发的安全与工程红线：登录认证、注入与 XSS/CSRF/SSRF 防护、越权、业务逻辑与竞态、数据库高并发、异步架构、接口防抓包改包与重放、密钥运

How this app is built, deployed, configured and kept alive on Data Clerk hosting (the data-clerk.com Kubernetes cluster). Use it whenever the user wants to deploy, publish, ship or "put live" the app, add or change an environment variable, API key or secret, edit secrets.env or use sops, install the tools for that, add a database table or migration, asks about the database, backups or restoring data, a domain or HTTPS, why the site is down or did not update after a push, the Dockerfile or the Gi

Raises the ceiling on a task before work starts, using the four kinds of ambition (speed, altitude, depth, range) from Felix Rieseberg's 13 October 2026 essay on being more ambitious with models. Use whenever someone commissions substantial work and the ask is pitched at the literal task rather than the outcome behind it: a tracker, report, letter, analysis, reconciliation, listing, plan, build or fix. Also use it on aim higher, be more ambitious, think bigger, what would an expert do, go deeper

Idiomatic use of the libraries and frameworks a project depends on: learning each one's model instead of reinventing the wheel or fighting the framework. Use before writing a general-purpose helper (retries, caching, pagination, concurrency, parsing, validation, date math); when building on or extending a library or framework (ORM, data-fetching client, UI kit, state manager, workflow engine, test runner); when a library, or the installed version of it, seems not to support what you need; when c

Run a repository's tests on Crucible (crucibl.sh) microVMs and act on the results. Use when asked to run the tests, check that a change passes, rerun failing tests, find out why CI or a PR check (`Crucible`) is red, handle flaky or quarantined tests, shard a slow suite, set up Crucible for a repo, or move GitHub Actions jobs onto Crucible runners (runs-on crucible-2vcpu-arm). Covers the hosted Crucible MCP server (run_tests, get_job, status, get_pr_check, actions_setup) and the `crucible` CLI.

Make blur and glass look premium in UI, and prove it's readable: Apple-grade frosted glass (a measured material with thickness levels, a lit rim and a tinted level), ready-made glass surfaces (sticky nav, sidebar/panel, dock, menu, dialog over a dimmed blurred page, toast, sheet), Liquid Glass refraction, progressive blur, blur reveals, glows and depth, with React, Vue, Svelte, Tailwind and native (SwiftUI, Flutter, React Native, Compose) guidance, plus contrast and frost checkers. Use this when

Operate Asana through asana-gate, the policy gateway that puts a declarative YAML policy in front of the Asana REST API for AI agents, as MCP tools (asana_*) and the asana-gate command line. Use this whenever a task touches asana-gate in any way: writing or fixing a policy (where/allow/deny scopes, constraints such as required members, title prefixes, rate limits and confirmations), running asana-gate commands, calling its MCP tools, reading its refusals (404 Not a recognized ID, 403 asana-gate:

Senior software-engineering partner for the whole code lifecycle — implementing changes, reviewing code and pull requests, assessing repositories, evaluating architecture and design proposals, finding and prioritising technical debt, planning and executing safe refactorings, security-focused code review (OWASP), improving tests, and writing ADRs. Use it whenever the user asks to review, audit, assess, refactor, clean up, harden or design code, or to implement a non-trivial change where maintaina

Estimates web tasks on Umbraco / ASP.NET Core sites (Razor, Vue or Nuxt front ends, Bellissima or AngularJS back-office extensions, uSync, Examine, Hangfire, multilingual and RTL) when an AI coding agent writes the code and a developer reviews it. Looks for a spec (spec.md, docs/spec) or feature plan (docs/plans) first and, if found, reads it in full and estimates per plan phase; if not, says so and gives a quick estimate of the visible scope. Reads the codebase, sizes the coding at agent speed,

Turn any research paper into a comic-magazine style PDF story (A4, ~8 pages) that explains the paper smoothly with characters, panels and speech bubbles.

Turn a product idea, MVP, or new feature into a scoped plan and an implementable backlog of vertical slices with acceptance criteria. Use this whenever the user wants to design, scope, shape, plan or "think through" an MVP, a new product, a new feature, an epic, or says things like "I have an idea", "let's plan this feature", "what should the MVP be", "break this down", "create a backlog", "create the tickets/issues/work items", or asks how to start building something. Works both inside a codeba

Show the user what their coding agent actually sees in them. Reads local Claude Code / Codex session history and git repos (read-only, nothing uploaded), then produces one six-field card — a title, three real numbers, one roast line, one truth they haven't noticed, one idea only they can build — as a terminal card plus a screenshot-ready HTML card. Use when the user runs /mirror, says "mirror", "what do you really think of me", "roast me", "what do you see in me", "what have you learned about me

Designs, builds, tests and improves Claude skills from a first idea to a packaged .skill file. Use this skill whenever a user wants to plan, spec, create, validate, test, benchmark, fix or optimise a skill, or asks why a skill is not being picked up. Triggers on phrases like "design a skill", "create a skill for", "help me build a skill that", "improve my skill", "my skill is not triggering", "run evals on my skill", "optimise the skill description", "package my skill", or any request to work ou

Use when building, editing, validating, or debugging generic Tracecat automations through Tracecat MCP, including workflow DSL/YAML authoring, table design, unique indexes, run-python Tracecat imports, agent presets, ai.agent or ai.preset_agent workflows, executions, validation, and workflow best practices.

Prepares a travel-expense reimbursement claim (החזר הוצאות נסיעה) for an Israeli university after a conference trip abroad: collects receipts from the trip's receipts folder and from email (flight and add-ons, travel insurance, hotel, registration, conference dinner, poster printing, ground transport), saves each as a named PDF, and builds the Hebrew right-to-left expense spreadsheet from its bundled claim template, converting to shekels at Bank of Israel representative rates (שער יציג) and comp