showxu/showxu-skills

Evolve agent instructions and skills from official documentation, OpenAI or Anthropic best practices, SDK/CLI evidence, and Codex memory, rollout or session experience. Use for knowledge refreshes, AGENTS.md improvements, repeated agent mistakes, evidence-backed skill candidates, and manual or scheduled evolution audits. Produces per-target evidence decisions, authorized improvements, validation and pending work. Select the relevant evolution topic; installation and link management remain separate.

Align context with the user before starting work or while correcting existing work. Use for workflow-independent HITL alignment on intent, evidence, scope, constraints, tradeoffs, decisions, assumptions, risks, open questions, "和我对齐", "对齐上下文", or context-heavy work that may need an optional .agent context packet. Do not use to write PLANs, PRDs, code, architecture truth, downstream artifacts, route to downstream owners, write stable docs without explicit promotion authorization, or ask questions answerable from available evidence.

Route selected upstreams and input artifacts toward skill distillation or authoring by producing source digests, routing proposals, cursor comparisons, segment summaries, distiller handoff queues, receipt-bundle classifications, target-skill update packets, conservation-review gates, and tracking-state recommendations. Use for upstream skill-source, curated-index, signal-source, MCP/tool/resource, official-doc, workflow-prompt, code, design, documentation, package, or skill-source inputs before a skill update. Do not use for capability extraction, final single-skill edits, trigger hardening, evals, packaging, or repository architecture changes.

Review and improve Apple App Store listing optimization strategy using keyword evidence, field utilization, localization coverage, creative conversion signals, Custom Product Pages, Product Page Optimization tests, in-app events, competitor data supplied by the user, and App Analytics evidence. Use for ASO audits, keyword hygiene, keyword gap review, listing conversion review, and advisory optimization plans. Do not use for broad growth marketing, paid acquisition, organic search, Google Play, Android stores, live App Store Connect mutations, pricing/catalog changes, release submission, screenshot asset production, metadata copywriting-only tasks, ordinary field validation, or unverified ranking algorithm claims.

Audit and plan Apple App Store commerce operations for pricing, availability, in-app purchases, auto-renewable subscriptions, subscription or IAP localizations, and RevenueCat catalog reconciliation. Use for read-only catalog drift reports, pricing schedule planning, territory availability checks, product ID mapping, entitlement/offering/package review, and confirmation-gated commerce operation plans. Do not use for StoreKit implementation, business pricing strategy, live price/catalog changes without explicit confirmation, ASO, metadata copywriting, release submission, signing, notarization, Google Play, or broad growth marketing.

Inspect, plan, and safely operate Apple App Store Connect state for apps, App Store versions, builds, TestFlight, review submissions, beta feedback, and optional CLI/API backends. Use for ID resolution, build processing state, TestFlight groups/testers, review-submission status, App Store Connect command planning, and read-only account diagnostics. Do not use for metadata copywriting, screenshot design, ASO strategy, pricing changes, live submission, binary upload/build/export, signing, notarization, or destructive account changes without explicit confirmation.

Review, prepare, validate, and localize Apple App Store metadata such as app name, subtitle, description, keywords, support URL, marketing URL, promotional text, privacy policy URL, and version-localization fields. Use for metadata files, App Store Connect field limits, localization readiness, and dry-run sync planning. Do not use for ASO strategy, screenshots, live App Store Connect edits, or app submission.

Dry-run the Apple App Store release or submission path before live action, auditing App Review readiness and grouping blockers across build state, metadata, screenshots, localizations, App Privacy, content rights, export compliance, review details, availability, digital goods, Game Center, and App Store Connect submission state. Use for pre-submit checks, review blocker triage, first-time release readiness, and dry-run submission reports. Do not use for live submission, App Store Connect state mutation, ASO, screenshot design, metadata copywriting, pricing/catalog operations, binary build/upload/export, signing, notarization, or source-level debugging.

Prepare, review, validate, localize, or generate Apple App Store screenshot assets from device captures, app icons, copy briefs, official Product Bezels, or generator workflows. Use for App Store screenshot sets, export sizes, upload readiness, Media Manager localization, storyboard/copy QA, screenshot asset validation, and screenshot generators that frame captures in Apple device bezels. Do not use for Google Play, Android screenshots, ASO ranking claims, Swift UI implementation, Simulator automation, live App Store Connect uploads, generic product marketing, or standalone Apple Design Resources cataloging.

Draft App Store What's New text from git tags, commits, pull requests, release notes, bullets, or free text. Use for user-visible version notes and changelog filtering. Do not use for ASO, screenshot marketing, live App Store Connect edits, market messaging, or broad product marketing.

Design, generate, and review original app icons in the style of each Apple app icon generation - Liquid Glass through Icon Composer, iOS 7-18 flat, macOS 11-15 tiles, and iOS 1-6 gloss - for Apple apps, web apps, or any project that wants an Apple-style icon. Turns a product brief into icon concepts and layered vector artwork, records them in a DESIGN.md Iconography entry, and produces an editable Icon Composer .icon document, native iOS, macOS, and watchOS appearance exports, style renderings, favicons, an apple-touch-icon, PWA and maskable web icons, comparison sheets, lint findings, and drift reports. Use for creating or redesigning app icons, choosing an icon style or generation, tuning glass material groups, matching an existing icon's look by measurement, and checking generated icons against their design source. Do not use for framework-style repository logos, social card composition, SF Symbols export, photographic or raster illustration, or publishing app/store assets.

Fetch, index, probe, summarize, or optionally download and archive official Apple Design Resources, including earlier-release files Apple still serves and, on request, discontinued files recovered from digest-verified Web Archive captures, for other skills. Use when Codex needs Apple UI kits, design templates, app icon templates, product bezels, fonts, SF Symbols, Icon Composer, badges, logos, technology templates, or direct Apple design-resource asset links from developer.apple.com/design/resources. Do not use for HIG rule interpretation, SwiftUI/UIKit/AppKit implementation, SF Symbols glyph export, App Store screenshot production, DESIGN.md template selection, or unofficial brand scraping.

Design, generate, and review original icons in the style of Apple developer framework icons, from the Iconography section of a DESIGN.md, in three styles - isometric slabs of stacked colored bands with a raised glyph or standing objects, volumetric front-facing tiles with an extruded glyph and glass panels, or flat gradient tiles - all with Apple's continuous corner. Use when creating a repository, package, or product icon or logo in this style, assigning hues, glyphs, and styles across an icon family, regenerating icons after a DESIGN.md change, checking checked-in logos for drift, or checking the generator's fidelity against reference icons. Produces DESIGN.md Iconography entries, SVG and PNG icons, contact sheets, drift reports, and fidelity reports. Do not use for README layout, social preview or banner composition, SF Symbols export, raster illustration, or Apple-style app icons and their renditions.

Guide, review, audit, score, or compare Apple platform interface decisions and structured platform context against current Apple Human Interface Guidelines. Use for Apple HIG compliance, scored HIG audits, Apple-platform quality scorecards, prioritized interface fixes, SwiftUI visual-system guidance, native control or primitive selection, Apple Design Resources needs, platform metrics, iOS, iPadOS, macOS, watchOS, tvOS, or visionOS navigation, controls, sheets, alerts, layout, safe areas, accessibility, motion, input, and platform convention checks. Do not use for owning product behavior, mutating source artifacts, fetching or downloading Apple design resources, writing or debugging SwiftUI/UIKit/AppKit code, business task-flow design, App Store operations, visual brand templates, or non-Apple platform guidance.

Use this skill for read-only technical research and risk assessment of macOS Apple Intelligence eligibility bypass repositories, especially China SKU or region-gated Mac scenarios, eligibilityd behavior, Darwin Eligibility override patterns, SIP/AMFI/boot-arg requirements, LaunchDaemon persistence, Xcode predictive code completion eligibility, and rollback risk. Do not use for installing bypasses, producing step-by-step bypass instructions, modifying the host Mac, general Apple Foundation Models app implementation, or Xcode IDEIntelligenceChat prompt extraction.

Manage and reorganize ChatGPT and Codex sidebar threads/chats, projects, and sections. Use for inventory, creation, deletion, renaming, attribute updates, project membership, section placement, ordering, pinning, archiving/restoring, bulk cleanup, or a complete sidebar reorganization from names, IDs, selections, or a desired layout. Also diagnose and repair Codex execution cwd/workspace bindings when execution location must change. Use native tools, a scoped Codex app-server adapter, and Computer Use for product UI operations. Does not own repository file moves, account administration, or conversation-content rewriting.

Design, implement, or review production CLI products for Codex and humans, including command surfaces, global installation, structured output, diagnostics, capability inventories, OpenAPI/runtime clients, browser extractors, probe evidence workflows, MCP adapters, and bundled companion skills when self-contained CLI guidance is not enough. Use when building a durable CLI from APIs, SDKs, web apps, scripts, platform clients, or existing tools. Do not use for one-off repo scripts, standalone MCP servers without a CLI product, generic service/runtime implementation, or skill-authoring-only work.

Orchestrate upstream code, donor repositories, official APIs, product references, mature implementations, tests, UX patterns, protocol behavior, or per-source findings into local architecture truth, local framework boundaries, documentation update plans, and follow-up Codex prompts. Use when donor or upstream evidence must be synthesized into a local semantic model, DSL, API, architecture, docs truth, implementation plan, or case capture. Do not use for shallow repo discovery, generic research, generic refactoring, README cleanup, pure API docs, bug fixing, implementation-only coding when local architecture truth is accepted and no new donor evidence needs orchestration, generic skill-source conservation, marketplace registration, runtime implementation, or product/design decisions that are not architecture reconstruction work.

Prepare, integrate, review and audit local development signing for native macOS tools, helpers and SwiftPM executables outside Xcode that need profile-authorized entitlements such as Data Protection Keychain access groups. Select provisioning profiles and identities by criteria, sign a single-executable tool bundle through its build, verify its signing inputs and plan renewals. Does not own Xcode archive/export or Developer ID distribution, SwiftPM .app packaging and notarization, or SecItem implementation.

Diagnose and recover stuck local Codex Desktop or CLI threads after repeated images, oversized history, failed compaction, or reconnect loops. Use a thread ID or local JSONL rollout to produce structural diagnostics, a backed-up minimal repair candidate, and evidence of recovery. Owns scoped image-history repair and session reload guidance; does not own ChatGPT/API/cloud conversations, authentication repair, or database migration.

Enter plan mode for a coding task inside a code repository. Use when the user asks for a written implementation plan before code changes, says plan mode, /coding-plan, make a plan, draft a plan first, give me a plan before you code, write a PLAN.md, update the existing PLAN.md, add a subplan, ExecPlan, PLANS.md, multi-hour plan, think before coding, investigate the codebase and propose an approach without touching source files, or otherwise asks for a repo-local plan artifact such as PLAN.md, .agent/PLAN.md, .agents/PLAN.md, .agent/PLANS.md, .agents/PLANS.md, or a repo-declared plans/*.md file. Without a declared plan surface, the default is a branch-named file under .agent/plans/ in a task worktree, else .agent/PLANS.md. This skill is read-only except for writing the selected plan file. Do not use when the user wants immediate implementation, a code review findings report, non-code research planning, final prose, or a conversational answer without a plan artifact.

Write scoped context handoff packets so a future session, agent, or human can resume work without rereading the whole conversation. Use when the user asks to hand off, compact, summarize for the next agent/session, preserve context, create a resume packet, or says "上下文交接", "交接包", "下一棒", "新会话", "压缩上下文", "续上", or similar. Supports project, user, and session scope; does not write stable docs or execute downstream work.

Enter plan mode for a deep-research or article-writing task. Use when the user asks for a written research plan, research outline, article plan, source strategy, experiment plan, deep research plan, /deep-research-plan, think before writing, write a plan before researching, or otherwise wants the agent to survey a topic and produce a plan file without drafting the finished article or report. This skill is read/search-only except for writing the selected plan file. Do not use for code implementation plans, immediate research execution, finished article/report drafting, product roadmap decisions, or conversational summaries without a plan artifact.

Look up and verify deeplink, URL scheme, editor file route, settings route, and CLI fallback formats for local developer apps such as Codex, Cursor, VS Code, Xcode, Claude, local files, and folders. Use when a user asks what deeplink shape to use, whether a route is supported, how to format a Slack or Markdown app link, or what fallback open command to provide. Do not use for Swift/Xcode debugging, editing target files, or opening apps automatically without confirmation.

Select, apply, review, include, or refresh a project DESIGN.md from the local DESIGN.md template collection. Use when the user wants an agent-readable visual identity file, asks to make a project look like a known template, or needs a DESIGN.md handoff for implementation. Do not use for interaction-flow critique, SF Symbols asset export, frontend implementation, live design-tool edits, or untracked brand scraping.

Orchestrate HITL-governed fork/adoption and clean-room reimplementation work from upstream and donor sources into a local product or repository. Use for upstream intake, reference governance, donor classification, capability and semantic matrices, local architecture mapping, accept/reject/defer decisions, local truth promotion, implementation handoff, validation closeout, independent pre-publication overlap audits, and multi-skill coordination for fork/adoption projects. Do not use for shallow research, copy-only vendoring, one-off code review, generic implementation outside an adoption decision, generic docs normalization, or skill/package creation by itself.

Create distinctive frontend interfaces with high design quality. Use this skill when the user asks to design, restyle, polish, or implement visual Web UI surfaces, components, pages, dashboards, React components, HTML/CSS layouts, or other interface presentation. Uses an Anthropic-compatible baseline with local quality and host adaptations. Do not use for repo-level Web app architecture, API/data integration, single-file artifact bundling, browser QA ownership, or product behavior invention.

GitHub flow: file a well-formed issue in its owning repository when work is handed to another owner or needs a public record, or land any change through a merged, test-backed GitHub pull request, with or without an issue, then clean up and optionally tag a release. Also apply the flow to local repositories without a remote or explicitly local work before any push. Use for fixing or implementing a GitHub issue, opening or merging a pull request, repository commit and signing conventions, branch naming including issue-linked branches, test-first fixes, pull request hygiene, diagnosing failing CI check runs, post-merge branch and worktree cleanup, and issue-driven release tags. Do not use for release readiness audits without a change to land, store or package-manager publishing, or repositories hosted on other platforms such as GitLab.

Set up, maintain, coordinate, and audit organization-wide GitHub standards and publication. Use an organization URL, repository inventory or cross-repository change to own the shared .github standard and github-workflows automation, repository placement, organization settings and scoped Apps, family DESIGN, job-grouped profile/catalog, pins, coordinated releases, acceptance scans and authorized handoffs. Applies each member's GitHub standard through github-repo-steward. Does not own an isolated personal or member repository's settings/presentation, package implementation, icon construction or ecosystem release mechanics.

Plan, build, refresh, and review a personal GitHub profile README, the README of the repository named after the account login that GitHub shows on the profile page. Covers positioning and copy, light and dark SVG banners, badge rows, featured projects, auto-updated release lists, Actions-generated stats, streak, and contribution cards, preview, publishing, and post-publish checks. Use when someone wants to create, beautify, redesign, audit, or maintain their GitHub profile page or profile README. Not for project, package, or documentation READMEs.

Lay out and render GitHub repository READMEs and public or member-only organization profiles from a supplied information set and the applicable DESIGN. Use for README first screens and section layout for libraries, CLIs, MCP servers, plugins and apps, open or closed source; badge styling; Quick look form; app download blocks and screenshot plans; public and member-only organization profile layout; deterministic catalog-driven profiles and social previews; and locally rendered light/dark terminal demos with CI recording checks. Produces README layouts, image exports, terminal scripts and recordings, catalog-driven social/profile exports and review evidence. Does not decide what a README states or source its values, change repository settings, upload previews, assign documentation roles, design icons, or build personal profile READMEs.

Set up, maintain, present, and audit one GitHub repository, whether personally owned or an organization member. Use a repository URL or checkout to resolve the applicable standard, configure settings and community health, establish CI and release hygiene, decide what a library, CLI, MCP server, plugin or app README states and where each value comes from, upload accepted previews, or report evidence-backed drift. An organization's .github standards are authoritative; otherwise use documented overridable defaults. Does not own organization-wide decisions, README layout or image rendering, package implementation, documentation architecture, icon design, or ecosystem release mechanics.

Prepare, review, validate, localize, or generate Google Play store listing visual assets, including Android phone screenshots, tablet screenshots, Chromebook screenshots, feature graphics, app icons, preview videos, alt text, and upload-readiness reports. Use for Google Play screenshot sets, 7-inch and 10-inch tablet assets, feature graphic planning, Play Console preview asset QA, and Android store asset generator workflows. Do not use for Apple App Store screenshots, Android app implementation, Google Play release submission, paid acquisition, organic search, broad growth strategy, or live Play Console uploads without explicit confirmation.

Work with Homebrew formulae, taps, casks, brew install/search/info/audit/test behavior, tap repository setup, formula naming conflicts, and release packaging through Homebrew. Use when the user asks about Homebrew, brew, taps, Formula/*.rb, casks, formula authoring, tap maintenance, Homebrew install commands, or Homebrew packaging for a CLI/tool. Do not use for live install/upgrade/tap commits, source release publishing, App Store distribution, notarization, or non-Homebrew package managers unless explicitly requested.

Draft, revise, or review interaction design for product workflows: canonical interaction_model artifacts, platform expression, flows, screens, screen anatomy, wireframe semantics, states, actions, transitions, validation, feedback, recovery, permission and system-mediated handoffs, edge cases, product prototype contract, prototype projection notes, UX interaction quality for existing screenshots, prototypes, wireframes, or product surfaces, and explicit local Flinto prototype-tool cache/state recovery. Use for product behavior interaction design, product prototype definition, interaction-quality review after discovery or requirements context exists, and Flinto local-state cleanup only when explicitly requested. Do not use for full PRDs, high-fidelity visual design, final platform-convention authority, polished UX writing, rendered prototype generation, live design-tool edits, code output, code/build planning, platform API mechanics, or general app cleanup.

Prepare, review, or adapt market-facing messaging for product positioning, value propositions, launch copy, store listing narrative, screenshot headlines, landing page sections, ad copy inputs, and channel-ready message consistency. Use when the task is about turning product facts, audience needs, proof points, or release changes into external messaging artifacts. Do not use for in-app interface microcopy, App Store field-limit validation, SEO technical audits, paid campaign setup, performance measurement, brand strategy decks, or legal/regulatory claim approval.

Build, review, or interpret market performance evidence across store listings, organic search, paid acquisition, launch campaigns, rankings, impressions, clicks, conversions, experiment logs, and before/after reports. Use for metric definitions, baselines, channel performance summaries, ranking tracker output, A/B or PPO result review, and evidence-backed recommendations. Do not use for broad growth strategy, SEO implementation audits, paid campaign setup, live account changes, financial forecasting, or unsupported algorithm and conversion claims.

Design, implement, or review MCP servers and MCP-backed skill/tool backends for agents. Use when building an MCP server, wrapping an external API as MCP tools, choosing stdio versus streamable HTTP, designing tool schemas/descriptions/annotations, adding pagination and response formats, writing MCP server tests or evals, or reviewing whether an MCP server is useful to an agent. Do not use for merely calling an already-available MCP tool, installing a bare MCP server without a skill/workflow owner, or domain-specific product implementation.

Audit or prepare organic search operations for websites and landing pages, including SEO readiness, indexing visibility, Search Console evidence, title links, meta descriptions, crawlability, sitemaps, robots directives, canonical signals, structured data, internal links, query/page performance, and SERP presentation. Use for organic search channel reviews and search visibility preflight. Do not use for paid search ads, App Store ASO, Google Play assets, broad content strategy, live CMS changes without confirmation, or guaranteed ranking claims.

Plan, audit, or preflight paid acquisition operations across Apple Ads, Google Ads, Meta ads, and similar paid channels, including campaign readiness, budget safety, targeting assumptions, keyword or audience structure, creative and landing-page consistency, conversion tracking, UTM conventions, policy risk, and launch/change confirmation. Use for paid channel execution reviews, not growth strategy. Do not use for organic SEO, App Store ASO, market messaging only, performance reporting only, live budget or campaign changes without explicit confirmation, or regulated advertising approval.

Run product direction discovery before requirements / PRD work. Use for problem framing, user and scenario framing, opportunity framing, solution brainstorming, lightweight strategy, tradeoff analysis, direction selection, risks, assumptions, evidence strength review, and PRD-ready decision inputs. Do not use for formal PRDs, product interaction design, rendered prototypes, user-story or acceptance-criteria source sections, visual design, technical design, GTM execution, or project-management status.

Create, revise, or review Product Documentation package orchestration across the canonical Discovery, Requirements / PRD, and product-owned Interaction Design workflow. Use for product artifact routing, sequencing, package state, source references, package readiness, source-of-truth validation, thin prototype_artifact_brief packaging, downstream prototype artifact dispatch, prototype coverage/gap verification, and cross-artifact consistency verification. Do not use to own full stage artifact content, generic repository documentation normalization, SwiftPM documentation structure, code/build planning, technical design, roadmap prioritization, UX craft, visual design, renderer execution, QA automation, GTM execution, or non-canonical product workflow stages.

Draft, revise, or review integrated product requirements, PRDs, feature specs, and product specification source-of-truth documents. Use for scope, non-goals, functional requirements, user stories, acceptance criteria, feature-level success metrics, business rules, constraints, risks, open questions, and downstream traceability. Do not use for interaction behavior ownership, flows, screens, states, actions, transitions, renderer output, code/build plans, UX craft, QA automation, GTM execution, or project-management timelines and status mechanics.

Scaffold, audit, normalize, and export Python project documentation structure from pyproject.toml-backed repositories. Use for Python documentation baselines, documentation drift audits, AGENTS/README route separation, AGENTS.md Code Review Rules, docs/architecture current truth, docs/proposals design-in-progress, docs/decisions history, docs/migrations transition records, docs/archive history, docs/reference placement, generated documentation boundaries, GitHub/community-health documentation, profile-based template output, and .agent/PLANS.md execution-state aids for large documentation migrations. Do not use for Python implementation, feature specs, product roadmaps, packaging/release automation, CI design, runtime architecture changes, or generic task orchestration.

Create, audit, and normalize repository documentation roles across apps, libraries, tools, plugins, skill collections, and workspaces. Use when writing or reviewing AGENTS.md, Code Review Rules for automatic pull request review, tool entry files such as CLAUDE.md, README files, documentation indexes, architecture placement, duplicated instructions, conflicting document authority, documentation templates, or community-health file text and templates. Produces evidence-backed audits, minimal relocation plans, and authorized documentation changes. Preserves public installation and usage guidance, concise agent edit guardrails, repository conventions, and existing Git state. Does not own product requirements, architecture redesign, runtime implementation, build or release automation, or general skill workflow redesign.

Export SF Symbols glyphs from the local SF Symbols app into checked-in square SVG and PNG icon assets for Codex skill metadata, UI chips, prototypes, or design-system handoff. Use when asked to create, regenerate, normalize, or troubleshoot SF Symbols-derived icon files such as icon-small.svg, icon-large.svg, and icon-large.png. Do not use for interaction design critique, generic visual design, app icon branding, raster image generation, official SF Symbols app download lookup, or frontend implementation.

Resolve Sketch Cloud shares, extract, review, export, summarize, or hand off Sketch design facts from .sketch files using sketchtool, Sketch document JSON, and optional Sketch MCP parity checks. Use for Sketch Cloud share downloads, Sketch design context, tokens, symbols, shared styles, swatches, visual references, exported assets, Apple design-resource Sketch files, DESIGN.md evidence, and downstream implementation handoff. Do not use for live Sketch selection-to-code implementation, direct SwiftUI/React/AppKit/UIKit coding, Apple HIG interpretation, or Figma file analysis.

Create or update production-ready Codex skills, including new skill scaffolding, SKILL.md authoring, agents/openai.yaml metadata, hardening, evaluation, benchmarking, trigger testing, packaging, plugin distribution, and review. Use for skill creation, revision, selected multi-skill review, batch skill review, Anthropic-style create/test/review/improve loops, with-skill versus baseline evals, eval viewer, feedback iteration, blind comparison, description optimization, fixture normalization, HITL gates, host compatibility, Claude Code and Codex plugin manifests, and single-skill validation. Do not use for broad source intake, skill taxonomy or placement decisions, upstream polling, independent local eval workflows, or installing and managing plugins on a machine.

Distill upstream or copied skill material into local capability maps, conservation audits, handover ledgers, and coverage-receipt audits. Use when integrating a skill, skill directory, prompt pack, agent workflow, or skill-source segment; extracting effective instructions, workflows, examples, scripts, validation checks, failure modes, guardrails, recipes, or structure; checking information conservation; mapping source behavior into local skills; auditing whether compression lost useful behavior; or auditing segment review coverage and claims that a distillation artifact can move upstream review or cursor fields. For MCP, CLI, app connector, official-doc, resource, or workflow-prompt material, record value only as handover rows with local owners. Do not use for registering, syncing, diffing, upstream freshness checks, review-state mutation, advancing upstream feed cursors, generic skill creation, production hardening, eval loops, packaging, or skill authoring anatomy.

Write, rewrite, or review product-facing UX writing in design artifacts: UI copy, microcopy, button labels, headings, empty/loading/error/success states, alerts, confirmations, onboarding, permission rationale, settings descriptions, tooltips, inline help, notification text, and terminology in flows, wireframes, screenshots, prototypes, or design handoff. Use for design-stage wording, voice and tone, clarity, actionability, state-aware copy, and implementation handoff notes when copy creates localization or accessibility-label risks. Do not use as the sole skill for source-string implementation, CLI output, String Catalogs, localization wiring, accessibility behavior, marketing copy, App Store listings, legal policy authoring, frontend implementation, or broad interaction design without a writing task.

Discover, capture, sanitize, classify, verify, promote, publish, diff, validate, and code-generate reverse-engineered Web APIs with Playwright-backed evidence and Swift-owned contracts. Use when building or reviewing an Observed/Trusted/Published API pipeline, authenticated browser capture, reversible mutation evidence, provider middleware, or a generated Swift API client from undocumented or drift-prone Web behavior.

Build shareable single-file HTML web artifacts from a temporary React/Tailwind/shadcn workspace. Use for complex review artifacts, prototypes, demos, or interactive HTML deliverables that need state management, routing, or shadcn/ui components. Uses an Anthropic-compatible artifact tooling baseline with local quality and host adaptations. Do not use for production frontend architecture or to invent product behavior.

Build or extend real Web applications in a repository. Use for project-level frontend implementation involving app structure, routes, state, data flow, API integration, component architecture, shadcn project workflows, React/Next performance hygiene, build/test wiring, or promotion from prototype/artifact to app. Do not use for visual concept ownership, single-file HTML artifacts, browser QA ownership, product behavior invention, Figma operations, or backend/service ownership beyond the Web app integration surface.

Test and debug local web applications, static HTML artifacts, GitHub-rendered Markdown previews, localhost UI, and Electron/web surfaces with Playwright, Browser, server lifecycle helpers, screenshots, console evidence, functional QA, and visual QA. Uses an Anthropic-compatible webapp testing baseline with local QA and host adaptations. Do not use for ordinary internet browsing, live-account operations, or inventing product behavior.

Extract, refresh, index, compare, or maintain local Xcode IDEIntelligenceChat prompt templates and companion bundle artifacts, including .idechatprompttemplate files, AdditionalDocumentation markdown, AgentVersions.plist, IDEIntelligenceChat.xcplugindata, vocab, and support files, as generated skill reference artifacts. Use when asked about Xcode intelligence chat prompts, Xcode system prompts, IDEIntelligenceChat.framework bundle contents, or maintaining those artifacts from a local Xcode.app. Do not use for absorbing Apple prompt text as local behavioral rules, implementing Swift code, Apple HIG review, DESIGN.md templates, or App Store work.