Claude Skills by Mixard

Use when implementing auto-renewable subscriptions, a free tier, or a bring-your-own-API-key option in an Expo/React Native app for iOS and Android, with server-side verification on Cloudflare Workers. Covers RevenueCat vs direct StoreKit 2/Play Billing, exact SDK call shapes, sandbox timing, App Store Server API and Play Developer API verification, and 2026 store policy.

Use when generating PNG icon imagesets (1x/2x/3x plus Contents.json) for Xcode asset catalogs, from SF Symbols (macOS, offline) or the Iconify API (275k+ open-source icons). Covers the bundled generator scripts, Iconify collection prefixes, and common SF Symbol names.

Use when implementing the iOS 26 Liquid Glass material in SwiftUI, UIKit, or WidgetKit. Covers glassEffect variants, GlassEffectContainer, glassEffectID/glassEffectUnion morphing, UIGlassEffect/UIGlassContainerEffect, scroll edge effects, and widget accented rendering.

Use when capturing a voice message in Russian, Ukrainian, or another non-English language in an Expo/React Native app and converting it to text on iOS and Android. Covers expo-speech-recognition, SFSpeechRecognizer vs iOS 26 SpeechAnalyzer/SpeechTranscriber, Android SpeechRecognizer/RecognizerIntent on-device models, and Gemini API audio transcription with pricing.

Use when preparing a consumer Android or iOS app for Google Play or App Store submission in 2026 - a dated compliance calendar (target API level, Play Billing, developer verification, AccessibilityService, foreground services, privacy manifests, Xcode/SDK minimums) plus a checklist for apps using AI-generated text, an accessibility overlay, microphone, chat screenshots, and subscriptions.

Use when writing or migrating Swift 6.2 concurrency code. Covers SE-0461/SE-0466 semantics (async stays on the caller, MainActor default inference), @concurrent offloading, isolated conformances, and Xcode 26 build settings.

Use when implementing or auditing accessibility on Web, iOS, or Android. Covers WCAG 2.2 success criteria (numbers, names, levels), contrast ratios, target sizes, and the cross-platform mapping for labels, hints, roles, and live regions.

Use when setting up or reviewing test/coverage/sanitizer tooling for modern C++ (C++17/20/23) projects — GoogleTest/CTest wiring, llvm-profdata/gcov coverage, and ASan/UBSan/TSan flags. Also covers a short list of C++ Core Guidelines rules whose correct answer contradicts a plausible guess.

Use when writing or auditing Solidity AMMs, LP vaults, or swap/deposit/withdraw flows. Covers vulnerable-vs-hardened pairs for reentrancy/CEI, donation-inflation share math, TWAP oracles, slippage and deadlines, SafeERC20/Ownable2Step/FullMath, and slither/echidna/forge fuzz commands.

Use when creating reproducible cross-platform (macOS/Linux) dev environments with Flox, or when a project mentions .flox/, manifest.toml, or flox activate. Covers the exact manifest.toml schema ([install], [vars], [hook], [profile], [services], [include], [options]), version pinning, pkg-group/priority conflict resolution, per-language recipes, and CLI commands.

Use when doing quick bioinformatics lookups across genomic reference databases with the gget CLI or Python package — finding Ensembl IDs, gene metadata, sequences, BLAST/BLAT searches, enrichment, or protein-structure references. Covers the module set, CLI/Python call shapes, exact flags, and install workflow.

Use when building database access with JetBrains Exposed 1.0.0. Covers newSuspendedTransaction, DSL vs DAO styles, upsert/batchInsert, a custom jsonb ColumnType, UUIDTable definitions, HikariCP and Flyway wiring, H2 test setup, and pinned Gradle coordinates.

Use when building Ktor 3.x HTTP servers. Covers the non-obvious defaults and gotchas — CORS's method allowlist, StatusPages nearest-class exception matching, JWT validate-returning-null semantics, WebSocket protocol details — plus JWT/StatusPages/CORS/WebSocket install-block shapes.

Use when writing or reviewing Perl 5.32+ code and unsure which feature is available in which version, or when porting legacy Perl idioms forward. Covers the version-gated feature table (5.32/5.36/5.38/5.40) and the legacy-to-modern anti-pattern mapping.

Use when working on any project built on the tinystruct Java framework — creating Application classes, @Action-mapped routes, HTTP/CLI dual-mode handling, JSON with Builder/Builders, database persistence with AbstractData, POJO generation, Server-Sent Events, file uploads, and MCP tools/servers. Covers exact APIs, annotation modes, and the dispatcher CLI that a model would otherwise guess.

Use when managing an Uncloud cluster with the `uc` CLI — deploying services, configuring Caddy ingress, adding static proxy routes for non-cluster devices, publishing ports, scaling, inspecting logs, or managing machines and volumes. Covers the CLI command surface, Compose extensions (x-ports, x-caddy, x-machines), port-publishing syntax, and internal service DNS.

Use when querying official US patent or trademark records — granted patents, pre-grant publications, application status, assignments, or trademark status via USPTO systems. Covers the PatentSearch (PatentsView) API v1 with its JSON query DSL and X-Api-Key header, TSDR trademark retrieval, ODP file-wrapper search, and the environment-variable auth conventions.

Use when writing or debugging E2E tests for Windows native desktop apps (WPF, WinForms, Win32/MFC, Qt) with pywinauto and Windows UI Automation (UIA). Covers per-framework UIA reliability, testability setup, the pywinauto locator/wait/action API, Qt version-specific gotchas (QT_ACCESSIBILITY, Qt5/Qt6 window class names), screenshot fallback with DPI rules, and Job Object process containment.

Use when writing modern Angular (v19-v21+) code - Signal Forms, linkedSignal, resource/httpResource, effect/afterRenderEffect, native CSS animations (animate.enter/leave), Angular Aria headless components, and the Angular CLI MCP server. Covers version-gated APIs and gotchas that changed after Angular 19.

Use when building a self-contained HTML presentation deck. Covers a viewport-safe CSS base with clamp() scaling and short-height breakpoints, slide density limits, a compact catalog of 12 visual presets (font pairings and palettes), and the negated-CSS-function gotcha.

Use when working on or reviewing Next.js 16+ projects. Covers the proxy.ts middleware rename (do not flag it as misnamed), Turbopack-by-default dev behavior, filesystem caching, and when to fall back to webpack.

Use when recording a polished UI demo/walkthrough video of a web app with Playwright. Covers cursor and subtitle overlay injection, moveAndClick/typeSlowly helpers, pacing values, video output handling, and recording gotchas.

Use when running dev servers or Python services under PM2, especially on Windows. Covers the .cjs config requirement, exact framework bin paths for Vite/Next/Nuxt, the Python child-process wrapper, and process-list persistence.

Use when writing or reviewing Prisma ORM code in TypeScript backends — schema design, query optimization, transactions, pagination, and serverless deployment. Covers critical traps like updateMany returning a count not records, the 5s interactive $transaction timeout, migrate dev resetting the DB, @updatedAt skipped on bulk writes, P-code error handling, and serverless connection exhaustion.

Use when asked to watch, audit, compare, or fix another agent's work given only artifacts - a session ID or transcript, a scheduled/cron run, a PR, branch, CI run, log, or pasted summary. Reconstruct what was asked, verify what happened against evidence, report gaps, fix narrowly only when authorized. Not for a task you just dispatched - that is subagent-driven-development.

Use before any creative or feature work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements, and design through dialogue before any implementation begins. When the design is settled, consider writing-plans next.

Use when asked to re-examine existing materials, plans, or results with fresh eyes and find what was missed - stale assumptions, blind spots, contradictions with own data. Russian triggers - "подумай", "посмотри под другим углом", "что мы упускаем", "критическое мышление", "что не так". Not for dead ends - if something is claimed impossible or blocked, use getting-unstuck instead.

Use before implementing, integrating, upgrading, or debugging anything that touches a third-party API, library, framework, CLI, cloud service, or provider SDK - and whenever the user asks for "latest", "current", "official", or "recommended" behavior. Forces reading current authoritative docs before writing code from model memory. Not for declared dead ends - that is getting-unstuck.

Use when you have a written implementation plan to execute inline in the current session, following it task by task with review checkpoints. After substantive tasks, refresh the project-cartography files if the project keeps them.

Use when a deliverable carries client facts - names, people, phones, emails, handles, domains, prices, codes, IDs, licenses, legal dates and thresholds, quotes - in drafts, emails, landing pages, reports, ad copy, configs. Also when the owner says "откуда ты это взял", "это неправильно", "такого нет", "не выдумывай". Placeholder over guess, facts file over memory, ledger check before delivery.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work. Guides completion by presenting structured options for merge, PR, keep, or discard, and handles worktree cleanup.

Use when about to conclude something is impossible, unsupported, or blocked - before reporting a dead end to the user. Especially after 2+ failed approaches, when a constraint comes from memory rather than a test, or when the phrase "can't be done" is forming.

Use when an agent must open files just to learn what they contain - a notes, research, or docs folder expensive to navigate. Generates a regenerable index, hand-writes preview headers only where the title hides the content, splits a data pipeline into input/output, refreshes drifted headers. Triggers - organize this knowledge base, annotate these docs. Not for source trees.

Use when one implementation plan should run in several Claude Code sessions or worktree subagents at once without conflicts: splits the plan into lanes with exclusive write-sets, a frozen contract layer, atomic git claims and a merge order. Triggers: 'в нескольких сессиях', 'параллельно в сессиях', 'run this plan in parallel'. Extends writing-plans; each lane executes via executing-plans.

Use when starting a project expected to outgrow one session, when joining a large unfamiliar codebase, or when a session begins in a project that keeps CODEMAP.md/PROJECT_STATE.md. Maintains a living three-file map (what is where, what is happening, why decisions were made) so each session starts from ~300 lines instead of re-reading the codebase.

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable. Requires technical rigor and verification, not performative agreement or blind implementation.

Use when the user wants a continuous, self-driving hunt for solutions to an open-ended problem - subagent generators, combination, adversarial verification with executed evidence, iterating until stopped. Russian triggers - "ищи варианты нон-стоп", "перебирай идеи", "не останавливайся пока не найдёшь". Not for existing materials (critical-review) or declared dead ends (getting-unstuck).

Use when long-running or parallel agent work must respect 5-hour and weekly usage limits - checking usage between waves of subagents, pausing near the cap with self-sufficient wake prompts, and resuming only when the window is clear. Pairs with any /loop-style autonomous run, including solution-hunter Stage 1.

Use when executing an implementation plan with mostly independent tasks in the current session and subagent dispatch is available. A fresh subagent implements each task, a reviewer gates each task, and a whole-branch review closes the work.

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes. Especially when under time pressure, when a "quick fix" seems obvious, or when previous fix attempts have failed.

Use when implementing any feature or bugfix, before writing implementation code. Also use when tempted to write code first and test after, or to skip tests "just this once".

Use when starting feature work that needs isolation from the current workspace, or before executing implementation plans. Ensures an isolated workspace exists via native tools (--worktree/EnterWorktree) or a git worktree fallback.

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - and at the start of multi-part work, to write the acceptance checks before implementing. Requires running verification commands and confirming output before making any success claims; evidence before assertions always.

Use when you have a spec or requirements for a multi-step implementation task, before touching code. Produces a task-by-task plan an engineer with zero codebase context can execute. For a fresh or large project, scaffold project-cartography first so the plan lands in a mapped codebase; execute via executing-plans.