Coding & Development

Quality-check translated software strings (English → Spanish by default) before they ship. Runs a script that catches broken placeholders ({{name}}, %1$s, %@, ICU plurals), lost HTML or XLIFF tags, changed URLs and numbers, strings over a character limit, missing or extra keys, untranslated strings, missing ¿ ¡, then reviews what a script can't see — register (tú/usted), terminology consistency, meaning, gender and plural agreement around placeholders, and locale formats. Use whenever someone shares localization files (JSON, iOS .strings or .xcstrings, Android strings.xml, gettext .po, XLIFF/.xlf) or asks to "check", "QA", "review", "lint" or "validate" a translation of an app, website, game or UI. Also trigger for "did I break any placeholders", "will this fit the button", "revisar la localización", "QA de strings". Not for translating from scratch (use translator-en-es) or for RAE rule disputes in prose (use rae-spanish).

Explains any concept as a Pokédex entry. The concept evolves through three stages of understanding (beginner, intermediate, expert), with type matchups for where it works and fails, a moveset of key properties, status conditions for misconceptions, and a gym battle quiz. Use when the user says "dex entry", "Pokémon style", "evolve this concept", or wants a clear, layered explanation of any topic in game style.

Explains any concept as a Souls-like boss fight (Elden Ring / Dark Souls style). The learner studies the boss's moveset (how it works), survives phase 2 (edge cases), and learns what kills most players (misconceptions). Use when the user says "boss fight", "souls mode", "explain like Elden Ring", or asks for a deep, edge-case-proof explanation of any topic in game style.

处理用户自己录的播客或口播录音（本机音频文件）：降噪、修剪首尾静音、人声均衡与压缩、线性响度标准化到约 -16 LUFS 并导出 MP3；可选片头片尾配乐和分段配乐（cue 表，出片前先预检）；给出长停顿收紧和重录口令标记段的剪辑建议，按用户确认的剪切表剪切。用户说「剪播客」「处理录音」「播客降噪」「响度标准化」「母带处理」「做 cue 表」「剪掉重录的地方」时使用。需要本机装有 ffmpeg 和 Python 3.9 以上，仅支持 macOS 和 Linux。不生成或克隆人声、不配音、不转写、不发布。

Runs a design-school style desk crit on a designer's own work (screenshot, image or Figma frame). Names 2–3 things that genuinely work, only the 1–3 most important issues with the principle behind each (hierarchy, contrast, alignment, rhythm, typography, color, composition), draws numbered markers on the work, assigns one time-boxed exercise for the main weakness, and suggests what kinds of references to study. Keeps a progress journal in the student's project folder so recurring issues are tracked and escalated across sessions. Use whenever a designer at any level (student, junior, self-taught, or experienced and asking to "be tough") asks for a crit, critique, feedback, review or "what should I improve" on their own design work, portfolio piece or course assignment. Not for reviewing someone else's product or decoding client feedback.

Give agents a namespace map of a codebase so searches start scoped instead of broad. Use to improve a codebase's token efficiency; save tokens and cut context cost when agents search code: at the start of work in an unfamiliar or large repo, when a task needs repeated codebase search, when a grep or glob returns a wall of hits, when results keep landing in unrelated parts of the tree, or when the user asks how to make their codebase cheaper or more token-efficient for agents.

ONE unified frontend design tool merging 11 capabilities into a single skill: designlang (extract any site's design system), taste-skill (anti-slop art direction), impeccable (24 design refinement commands), GSAP (advanced animation, 8 modules), transitions.dev (43+ production CSS transitions), react-bits (218 animated React components), playwright-cli (browser automation & testing), plus 4 original modules: rtl-arabic (Arabic/RTL interfaces), a11y (WCAG 2.2 AA accessibility), design-tokens (DTCG token architecture & exports), and web-vitals (Core Web Vitals performance). One installation, one entry point, one workflow. Use when the user wants to: extract a design system from a URL, build/redesign/polish/critique/audit a frontend, add animation or transitions, use animated React components, automate/test a browser, build an Arabic or RTL interface, make UI accessible, create design tokens, or make a page fast/pass Core Web Vitals. Triggers: 'extract design', 'design system', 'build a landing page', 'redesign

Operational advisor of the Digital Lean Framework (DLF v2.4), in English. Guides DLI assessments with the Efficiency Matrix, 9 Muda, three sources of inefficiency, Sprint Kaizen, Law Zero, Three Laws of Flow (Little, bottleneck, variation), digital Jidoka, internal customer-supplier chain, PULL and Standardization. Applies to 4 operational tracks (IT and Digital Service Delivery, Cybersecurity, Leadership & People, AI Adoption) + 1 transversal layer (Digital Transformation Programs). Use for: lean assessment, muda scan, value stream mapping, DLI, kaizen, flow ratio, complexity score, value alignment, remediation, jidoka. Triggers with: "Lean Advisor", "assessment", "muda", "DLI", "kaizen", "waste", "law zero", "value stream", "WIP", "bottleneck", "matrix", "Ocean", "Islands", "jidoka", "stop the line", "operational track", "transformation program", or operational references to the DLF. If in doubt, trigger. NOT for generic business questions.

Publish Markdown files to shareable URLs and convert them to branded PDF or DOCX files using the OpenDocs CLI. Use this skill whenever the user wants to publish, share, or export Markdown — including asks like "turn this into a shareable link", "give me a PDF of these notes", "upload all the docs in this folder", "update the post I published", or "convert this .md to Word". Also use for finding and reading the user's published documents to answer questions about them ("what does our NDA with Acme say?", "summarise the Q3 reports"), for managing published documents (listing, updating, unpublishing), for saving documents into the workspace's Google Drive folders or sending them for e-signature, for any workflow where Markdown needs to become either a URL or a formatted document file, and for connecting Claude, ChatGPT or coding tools to the OpenDocs MCP server.

Your brand designer. Names the business (with the trademark, domain and handle checks to run before falling in love with a name), writes the voice, a one-line promise and a tagline, and briefs the look: colors, type, logo and the first things a customer sees, so they match the positioning the marketing director chose. Use when the user says "name my business", "brand this", "logo", "brand identity", "what should it look like", "tagline", or "brand voice".

Create or extend a visual identity with whatever providers the user has: brand brief, color palette, type pairing, logo concepts (vector SVG authored directly, plus generated explorations), and applied mockups (social posts, profile images, banners, merch, packaging), saved as a reusable brand kit that other creator skills read. Use when the user asks for a logo, brand identity, brand kit, color palette, style guide, channel branding, or consistent branded social templates. Not for single product photos (creator-product-photo) or thumbnails (creator-thumbnail).

Use on Day 1 of a Brand OS install, or when someone says "configura mi marca", "onboard me", "iniciar marca", "set up my brand", or wants to populate the Brand Brain. Runs a 7-block interactive interview and scaffolds the canonical Brand Brain in vault/Conceptos/.

Turns a short product brief into a one-page brand system (name, colors, type, tagline, voice) saved as brand-system.md, ready to paste into Lovable project knowledge.

Keeps the brand's look on file in Brand Assets — builds and maintains the Brand DNA doc (typefaces, colours with hex, photography direction, product details, ad style) from one read of the brand's own site and the photos already kept, keeps the product photos and the logo the user uploads (one folder per product inside Brand Assets), judges whether an unnamed upload is a brand asset at all, and adds a new product's details to the doc. Use when the user asks to build, fill, fix, correct, update or rebuild Brand DNA or the brand's look; uploads or sends product photos or a logo, named or not; or when another skill finds Brand DNA empty or needs an upload kept. Not for making ads — static-ads makes them and only reads what this keeps.

Gives a product you will sell its own identity, and scans the code for traces of the original. Two phases: identity early (name candidates with the trademark and domain checks to run, palette, voice, logo brief) and a sweep before launch that searches the code's text files for the original's name, domain and colours, including inside identifiers, and blocks deploy until it finds none. The sweep is a bounded token scan, not proof that nothing of the original remains: images, fonts, hsl() and named colours need a look by hand. Use for "name this product", "rebrand the clone", "check for leftovers of the original", "brand sweep", "is it different enough". Full (sell) track of the swap series. Do NOT use to build the design tokens, use swap-design. Do NOT use to write the landing page or store listing, use swap-launch.

Batch-land open Renovate PRs in the current repo. Enumerate Renovate-authored PRs, classify each as GREEN / CONFLICTING / REAL_FAILURE / HOLD, approve and rebase-merge the safe ones in ascending PR-number order, nudge conflicting ones with the rebase checkbox, and stop cleanly on real failures or cross-file HOLD cases. Use when the user says "/land-renovate", "land the Renovate PRs", "clear the Renovate queue", or wants to merge a batch of dependency-update PRs.

Reinstall a user's DeepSeek Harness plugin set on a new machine from a version-pinned manifest, using the Desktop-aware CLI, optional profile config, and version-pinned local patches. Use when the user says they are on a new machine, want their DSH plugins restored, or want their current DSH setup reproduced elsewhere.

Interactive skill assessment with personalized learning path generation

用《小岛经济学》的生产率、储蓄、贷款、货币、贸易与危机分析回答和规划问题；先识别本书模型与作者立场，再给出可检查的条件和边界。

Use when a user is clarifying life direction, weighing a career or relationship choice, feeling stuck between values and action, or reviewing a setback. Chinese cues include 人生方向、职业选择、价值与行动脱节、反复纠结、受挫后怎么调整; English cues include life direction, career choice, values, decision, and setback. Not for diagnosis, crisis response, or making irreversible choices for the user.

Assess and add a CONVENTION (a reusable rule, practice, or naming/process standard) to a documentation-led repo — decides FIRST whether it is worth codifying at all, then routes it to the right home (AGENTS.md hard rule, CONVENTIONS.md guidance, GLOSSARY term, or to /new-adr if it is really a one-off decision). Pushes back on premature or duplicate conventions. Use when the user says "add a convention", "make this a rule", "document this practice", "we should always X", or invokes /add-convention. NOT for recording a single architectural/product decision (use /new-adr) and NOT for queueing work (use /new-plan).

Run a bounded wave over the plan/todo queue in a documentation-led repo, using available parallel agents or sequential execution. Establishes width, budget and supervision; isolates work and collects verified outcomes. Use for "spawn a wave of agents", "run the queue in parallel", "fan out the work", or "agent wave", not for authoring a decision or queuing a new item.

>- Закрытие параллельных сессий Claude Code в проекте и доводка хвостов: по каждой сессии проверить, что работа закончена, закоммичена, слита и записана; спасти материалы из временных папок; довести незавершённое; собрать список «за владельцем»; закрыть лишние сессии и оставить проект в законченном виде с одной страницей состояния. Используй, когда владелец говорит: «закрой лишние сессии», «вычисти сессии», «подбери хвосты», «убедись, что всё закоммичено и записано», «приведи всё в законченный вид», «переходим к следующему этапу — что осталось?», «куча сессий, я потерял нить», «собери всё в одну сессию», а также перед передачей проекта другой сессии, другому аккаунту или исполнителю. Не для закрытия одной своей сессии и не для обычного «закоммить».

Daily focus planner. Gathers work from Teams, Outlook, GitHub and Jira, ranks it into a focus list, and keeps a local HUD updated through the day. Use when the user says "/today", asks what to focus on, wants a morning brief, a midday refresh, an end-of-day wrap, or asks why an item is ranked where it is.