Explore

Manually run the extended DeepSeek Harness bilingual-document workflow, including generated briefings, delegated prose translation, whole-document translation, and scoped pairing verification.

Use when investigating or optimizing DeepSeek Harness performance, designing realistic synthetic benchmarks or CI performance gates, profiling long Sessions or Web responsiveness, or turning performance PR evidence into measured behavior-preserving fixes.

Use when writing, reviewing, restoring, trimming, or auditing prose in the deepseek-harness repo, including deciding where documentation or comments are required across Markdown, JSDoc, code and test comments, prompts, descriptions, diagnostics, and CLI or UI strings.

Use before pushing, force-pushing, marking ready for review, or claiming checks pass on a deepseek-harness branch, and immediately after gh stack sync publishes rewritten branches, to select the smallest tests and checks that cover the outgoing or just-published diff without reflexively running the full repository suite.

Use when landing a stack of dependent GitHub PRs (A ← B ← C, where each bases on the one below) onto master, merging a PR whose base is another open PR's branch, or whenever a request mentions "stacked PRs", "PR stack", "dependent PRs", or merging several related PRs in sequence. Requires every same-repository dependency chain to use GitHub's official stacked-PR feature before landing so GitHub owns stack-wide rules, CI, ordering, retargeting, and merge state.

Find evidence-backed simplifications in DeepSeek Harness code, APIs, configuration, tests, and prose; write or consolidate proposals, identify small inline cleanups, or assess simplifications from another branch. Use for removing dead, duplicated, speculative, or unnecessarily maintained behavior and infrastructure.

Create, restructure, review, audit, or migrate DeepSeek Harness Markdown documentation, package READMEs, and the documentation website using audience-first hierarchy, kind-mapped YAML metadata, bilingual line alignment, summary/contents navigation, progressive user-to-developer detail, executed-operation fact-checking, and repository validation. Use for new or revised DSH docs, docs-tree organization, documentation-quality audits and budgets, website page publishing, and bilingual documentation structure changes.

Use when reviewing a pull request in the deepseek-harness repo — orients the reviewer to this codebase's standards (AGENTS.md conventions, defensive patterns, ADRs, quality gates) and the review-specific checks that code alone can't show

Design and review DeepSeek Harness client UI changes — visual token discipline, reuse-before-adding, feedback surfaces (toast vs in-place notice vs empty state), overlay and menu safety, platform window adaptation, loading states, and error copy placement. Use when adding or changing product-user-visible GUI behavior in packages/client, or when reviewing such a PR.

Design, review, and diagnose DeepSeek Harness tests and fixtures that can fail nondeterministically under CI concurrency, shared host resources, clocks, process-global state, subprocesses, network listeners, or asynchronous teardown. Use when adding or changing tests with those risks, investigating flaky CI, or reviewing test isolation; use dsh-pre-push-checks separately to select outgoing commands.

Use when adding, auditing, pruning, archiving, restoring, or reviewing Agent Notes in deepseek-harness; checks every new note for superseded active records, deletes small UI and purely mechanical records, classifies other implemented notes by future decision value, deletes rejected notes that no longer prevent a tempting fallacy, and applies the frozen archived/{kind} triplet and manifest rules.

Resolve which slide, page, and (optionally) selected element the user is currently viewing in the open-slide dev server. Consult this whenever the user references "this page", "this slide", "this element", "the slide I'm on", "the current page", or any deictic reference to slide content without naming it. Re-read `node_modules/.open-slide/current.json` at the start of every such turn — the user navigates between turns, so a value you read earlier in the conversation is almost certainly stale.

Writing, exploit: shape raw material into an article, paragraph by paragraph.

Writing, explore: mine raw fragments, no structure yet.

Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.

Writing, exploit; assemble raw material into a journey of beats, grounding each term before a beat leans on it.

Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.

Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.

Stop. That last message did not land: re-pitch it.

Only use when the user explicitly invokes this skill by name.

Move issues and external PRs through a state machine of triage roles, categorise, verify, grill if needed, and write agent-ready briefs.

Make a fully code-generated animated video at the highest achievable visual and design quality (with an original synthesized soundtrack) whose look, sound, editing and pacing are learned from a reference video — in whatever style that reference uses (flat 2D, 3D/CG, anime, painterly, pixel art, hand-drawn, kinetic type, stop-motion feel…). Studies the reference frame-by-frame and by audio analysis, writes a measured style bible, derives an original story, picks a rendering approach that fits the style, builds a deterministic browser animation, renders it to MP4 with headless Chrome + ffmpeg, and QA-reviews it. Use this whenever the user shares a video/YouTube link or clip and wants an animation, short, trailer, teaser, explainer, mascot or character video "in the style of", "like this", "inspired by", or "with the vibe of" it — or asks you to study a video's art style, sound, camera, transitions or narrative and then make something similar — even if they never say "reference" or "skill".

Package an existing interactive course generated by the book-to-course skill (a folder with index.html, assets/ and course/data.js, or its .zip) into a signed Android APK — a light offline WebView app to install on a phone or tablet. Checks the build tools on the machine (JDK, Android SDK build-tools and platform), tells the user exactly what is missing and where to get it, and can download it for them. Use this whenever the user wants a course "na telefon/tablet", "na Androida", "jako aplikację", "zrób z kursu APK", "course to apk", wants to learn from their generated course on a mobile device offline, or wants to rebuild/update a course app after adding chapters — even if they don't say "APK" explicitly. Do NOT use for building a course from a book (that is book-to-course), for iOS apps, or for general Android app development.

Turn a book or document (EPUB, PDF, HTML, DOCX, Markdown, TXT — e.g. a programming course book) into a complete interactive beginner-friendly course delivered as a local web page, with step-by-step explanations, animations, quizzes, chapter tests, flashcards, hints, programming exercises checked by unit tests, and learner progress saved to files. Use this skill whenever the user uploads or mentions a book/ebook/PDF/EPUB/tutorial/lecture notes and wants to LEARN from it, wants a "kurs interaktywny", "kurs z książki", "zrób z tego kurs/naukę", an interactive course, study site, quiz/exercise version of a book, or says to continue/extend an existing generated course ("następny rozdział", "kontynuuj kurs"). Use it even if they don't say "skill" or "course" explicitly but describe turning reading material into something to study interactively. Do NOT use for plain summaries, a single quiz, or translating a book.

Use when creating, revising, or reviewing explanatory data charts for a blog, report, paper, slide, newsletter, or social post. Includes matplotlib, seaborn, pandas .plot, ggplot2, Plotly, Altair/Vega-Lite, D3, and PNG/SVG charts. Also use for "critique this chart", "why does this plot look bad", or "make this figure better" with a chart image. Not for unrelated coding, software architecture diagrams, or non-data illustrations.

旅行 vlog 地图视频：用户给一组旅行照片（和一张自己的照片），自动按照片 GPS 和拍摄时间找出机位、排好顺序、沿真实步行/驾车路线连起来；用照片生成用户的 Q 版卡通形象，卡通小人沿路线一颠一颠地走，右上角里程累加，走到每个机位停下「咔嚓」拍照、照片啵地弹出（写着第几站、已走多少公里），再走向下一站，最后照片墙 + 总里程收尾；配原创欢快 BGM 和脚步、快门、啵、叮音效，弹照片踩在拍点上。手账风竖屏 MP4。用 map-motion 引擎。当用户说「旅行 vlog」「把旅行照片做成视频」「小人在地图上走」「卡通形象走路线」「打卡路线视频」「citywalk 视频」「一日游路线动画」，或丢来一组带位置的旅行照片要做成视频时使用。单张照片找机位用 photo-spot；3D 山体用 map-3d。

找机位 / 机位揭秘视频：给一张照片，做一段竖屏短片——卫星地球俯冲 → 落出拍照机位（相机图标）和被摄地点 → 取景扇形从机位扫向被摄地 → 取景框对焦 → 咔嚓快门声 → 照片从机位处翻折立起（3D 透视），背景是这张照片的虚化；可加 iPhone Live Photo 播放效果（真 MOV 或静态模拟）。用 map-motion 的渲染引擎。当用户说「找机位」「机位揭秘」「这张照片在哪拍的」「把照片做成地图视频」「照片定位动画」「拍照地点动画」「咔嚓翻出照片」「Live Photo 效果」，或丢来一张风景/旅行照片要做成短视频时使用。

Screen + voice → structured work. The user records their screen while talking (demoing their product, browsing, exploring an idea); turn the recording into items — bug/polish issues with repro steps, frames and suspected code, ideas with the inspiration behind them, notes, todos — let them review on a local page, then export (Feishu/Lark Bitable, CSV/Markdown, GitHub Issues…) or start fixing. Use when the user wants to record feedback / QA / ideas by talking while using their screen, hands over such a video, or wants to process recordings from the Blurt app. Triggers: "blurt", "/blurt", "开始口喷", "口喷鸡", "开始录", "录屏提 bug", "录一下我的想法", "record feedback", "turn this recording into issues", "处理我录的视频", "process my recordings".

给视频配音：把旁白和角色台词合成成音频文件，标好时长，交给做视频的流程对齐字幕。做视频（make-video）、改视频（revise-video）时调用，也可以单独用来试音色。配了通义（阿里云百炼）的 key 就用通义，没配就用 macOS 自带的中文语音，都不需要工作台配置。

做法：原片剪辑（一键换装、舞台混剪）。画面全是同一首歌多个场次的真实直拍，声音是歌曲原声，按拍子换场、贴合换装。系列或内容用这种做法时，调研、写脚本、生成视频、修改视频各环节读这里对应的文件。

做法：解说视频（HyperFrames 动态图文 + 配音，系列配了角色时加解说员）。没有指定其他做法的内容默认用它。写脚本、生成视频、修改视频时读这里对应的文件；调研按流程技能的通用做法。

Creates showreel-grade motion graphics videos entirely from code — HTML scenes rendered frame by frame in headless Chrome with real motion blur, plus an original score composed for each video on the same beat grid. It directs the film itself from whatever the person gives — their words, a site, a GitHub repository, an app, screenshots, their own video clips and photos, or reference clips — and sets the pace and the energy of the music from them. Use for any promo, ad, launch video, explainer, reel, Shorts or TikTok, intro, kinetic type or logo animation for a business, product, app, site, bot, channel or person, even when all you have is a link or a name.

Use on Windows for unexpected DSH sandbox access denials: workspace writes or listing fail, or an ordinarily readable path cannot be read. One bundled command inspects the path and every ancestor and repairs the ACL problems it proves in that same run. Expected confinement denials need no ACL repair.

Record browser or Web UI interaction demos as optimized GIFs using the available browser-control workflow, optional Playwright Videos for higher capture frame rates, and deterministic encoding, then attach the GIF to a pull request with `gh --attach`, falling back to a dedicated assets branch where attach cannot apply. Use when asked to make, record, or generate a GIF that demonstrates a browser workflow, and for every pull request that changes product-user-visible GUI behavior, which MUST include a GIF recorded from the pull request's real server and model flow.

Stabilize a system that is failing right now — name the signal that flagged it, size the blast radius in numbers, keep a timestamped log written as you go, find the last known-good state, then propose the smallest reversible mitigation and confirm recovery against that same signal. Use when production is degraded or down and time to mitigation matters more than a complete explanation. Not for a defect that is not currently failing (systematic-debugging), not for the root-cause investigation or the postmortem that follows, and never authorization to touch production without asking first.

Make a layout hold from 320px to wide desktop — content-driven breakpoints, intrinsic layout before media queries, fluid type that still zooms, touch targets, reflow without hiding content, and what to check at each width. Fires when building or fixing layout, when something overflows or collapses at a size, or when a design only exists at one width. Not for proving it renders (browser-verification) and not for semantics or screen readers (accessibility).

Establish a visual direction for new or reshaped UI so it reads as designed for this product rather than assembled from defaults — hierarchy, typography, colour, spacing, density. Use when building a new surface, when something looks templated and needs a direction rather than a tweak, or when asked to make an interface feel considered. Prefer the official Anthropic frontend-design skill when it is installed; this is the local counterpart for when it is not. Not for critiquing an existing interface, and not for work inside a design system that already decides these things.

Design or change an HTTP API — resources, verbs, status codes, one error shape, pagination, versioning, and an OpenAPI description that cannot drift from the handlers. Use when adding or reshaping endpoints, when asked what a response should return, or when reviewing whether a change to a published API breaks its consumers. Not for choosing between REST, GraphQL and RPC, not for database schema design, and not for implementing the handler's business logic.

Decide what actually occupies the model's window — progressive disclosure through an index, retrieval versus inlining, what compaction must preserve, and what loading everything costs. Use when a prompt or agent pulls in many files, docs or tool transcripts, when quality decays over a long session, when choosing between fetching at runtime and pasting up front, or when a context has to be trimmed or summarized. Not for the wording of the prompt, not for an agent's job and tool boundaries, and not for picking retrieval infrastructure.

Scope an agent or subagent before it is built — the one job it owns, the smallest tool set that closes that job, what it must never do, and the evidence it has to return. Use when adding an agent, subagent or automated role to a system, when deciding which tools it gets, when an existing agent loops, over-reaches or reports work it did not do, or when reviewing someone else's agent design. Not for wording the prompt itself, not for deciding what occupies its context window, and it never grants an agent permission it did not already have.

Review PyTorch pull requests for code quality, test coverage, security, and backward compatibility. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".

Audits, creates and improves Agent Skills (folders with a SKILL.md) against Anthropic's skill authoring best practices, the Claude Code skills docs and the prompting guides for newer models such as Claude Fable 5 and Opus 5.5. Audit mode reports every rule as pass, fail or not applicable, with file and line evidence and a proposed fix, and changes nothing until the user approves. Use when the user asks to audit, review, check or score a skill or a folder of skills, to create a new skill or turn a repeated workflow into one, to improve, shorten, refactor or update an old skill for newer models, or to run the skill validator.

End-of-session handoff. Distils and archives the previous handoff, records what was done, what still runs in the background, what is next (the task of the day) and what waits for the user, in notes/handoff.md, BACKLOG.md, questions.md, CLAUDE.md, the topic notes and memory, then commits and pushes per the project's handoff rule, so any session on any machine can start from an empty context. Use it whenever the user wants to finish, close or wrap up a session ("handoff", "end of day", "save where we are", "switching machines", "battery dying"), before a /clear, or when the conversation is about to be compacted after a long stretch of work.

Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).

Turn the current conversation into a spec and publish it to the project issue tracker: no interview, just synthesis of what you've already discussed.

Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

Teach the user a new skill or concept, within this workspace.

Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.

Wire dependency-cruiser into a TypeScript repo so each package is a deep module, with implementation hidden in subfolders and reachable only through its entry-point files. User-invoked.

Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.

Configure this repo for the engineering skills: set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.

Create exercise directory structures with sections, problems, solutions, and explainers that pass linting. Use when user wants to scaffold exercises, create exercise stubs, or set up a new course section.

Use when you need to resolve an in-progress git merge/rebase conflict.

Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.

Migrate test files from `as` type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data.

Grill me about specs for the workflows I want to build, within this workspace.

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

Implement a piece of work based on a spec or set of tickets.

Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.

Use when building, styling, or reviewing in-harness UI — the BotHarness client plugin inside the DeepSeek Harness Web Client. Trigger on new or changed components under `packages/client`, layout or padding alignment with native DSH, tokens/primitives or theme questions, slot and shadow wiring, and any "match the native shell" request.