Skills de Claude Code · página 33

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Learn how SKILL.md files work in Claude Code plugins, then build a production-quality agent skill from scratch. Covers frontmatter schema, body structure, testing, and iteration.

Step-by-step guide to writing a SKILL.md file for Claude Code. Learn how to plan, structure, and test auto-activating skills with proper frontmatter, allowed-tools, dynamic context injection, and supporting files.

|

|

|

|

|

|

|

|

|

Use when applying Hyperflow's orchestration doctrine in Codex, Antigravity, or another single-agent surface. Auto-invoke for non-trivial engineering work: build, implement, add, refactor, debug, fix, review, audit, plan, scope, design, brainstorm, ship, or deploy.

|

|

|

|

|

|

|

Hyperflow prompt amplifier. Use when a prompt is rough, vague, or under-specified and should be rewritten to high quality before running it — "enhance this prompt", "make this prompt better", "improve my prompt", "rewrite this prompt". Domain-aware: detects the prompt's domain, injects the matching specialist standards + project rules, scores against an 8-dimension rubric, then offers to run it.

Hyperflow code review. Use when the user wants the current diff, a commit, branch, or PR reviewed — verbs like audit, review, "check for issues", "security check", "code review". Multi-level review (L1 quick → L5 exhaustive), writes findings to .hyperflow/audits/, then a fix-gate.

Hyperflow memory manager. Use to view, search, add, edit, prune, or clear hyperflow project memory — "show memory", "search memory for X", "clear memory", "what does hyperflow remember about Y". CRUD over .hyperflow/memory/ only — never touches source code.

Hyperflow ship phase. Use when the user is ready to release — verbs like ship, push, release, deploy, "cut a release", "ready to push". Runs pre-push gates (lint + typecheck + build + tests + security sweep), then asks before pushing. Never --no-verify, never force-push to main.

Hyperflow execution phase. Use when a task file exists in .hyperflow/tasks/ and the work needs building — verbs like build, implement, add, refactor, "wire up", "run the plan", "execute the task". Works batches sequentially with self-review and per-task commits. In Antigravity there is no sub-agent fan-out — the single agent does each batch itself.

Hyperflow project setup. Use when starting hyperflow in a new project or refreshing its cache — "init hyperflow", "set up hyperflow", "refresh hyperflow", "scaffold hyperflow". One-shot setup of the .hyperflow/ project cache + memory and the .agent/workflows/hyperflow* slash commands. Does not start the spec → scope → dispatch chain.

Hyperflow planning phase. Use when a task is clear enough to decompose into batched steps before writing code — verbs like scope, decompose, "plan out", "break down", "plan this". Read-only with respect to source; writes a task file to .hyperflow/tasks/<slug>.md, then hands off to hyperflow-dispatch.

Hyperflow design phase. Use when the user is exploring an idea, weighing approaches, or has an ambiguous request — verbs like brainstorm, design, explore, "should we", "what's the best way to", "unsure about". Thinking, not building. Produces an approved design at .hyperflow/specs/<slug>.md, then hands off to hyperflow-scope.

Hyperflow project status. Use to see current hyperflow state — "what is hyperflow doing", "show task progress", "where are we". Read-only — reports in-flight tasks, memory count, and progress. Never modifies state or runs work.

Hyperflow auto-routing mode. Use to control how aggressively hyperflow auto-routes — "make hyperflow sticky", "stop using hyperflow", "auto-route to hyperflow", "disable hyperflow auto-routing". Sets on (every task-shaped message routes) / auto (intent-verb messages route — default) / off (no auto-routing).

Hyperflow debugging. Use for bugs, test failures, runtime errors, broken builds, or "this doesn't work" reports — verbs like debug, "fix it", solve, "why is X failing", "Y is broken", or a pasted stack trace. Systematic root-cause analysis before any patch — never blind-patch symptoms.

Autonomous app builder that converts plain-English descriptions into

Developer relations playbook builder — produces a DevRel program design covering community platform, contributor program tiers, ambassador criteria, event strategy, and success metrics. Use when asked to "build a DevRel program", "design our developer community", "create a contributor program", "start a developer advocacy program", or "how do we grow our dev community".

Hacker News post crafter — given a product, feature, or story produces a ready-to-post HN submission (title ≤80 chars, no marketing), honest body text with technical depth, no outbound links, predicted reception analysis, and comment-response templates for likely pushback. Use when asked to "write an HN post", "craft a Show HN", "prepare our Hacker News launch", or "help me post on HN".

Media and podcast outreach personalizer — takes a story angle and target journalist or host list and produces personalized pitch emails per target. Use when asked to "write media pitches", "pitch this story to journalists", "get us on podcasts", "write press outreach", or "personalize pitches for these contacts".

Content brief generator — takes a topic or keyword and produces a complete content brief with target keyword, search intent, recommended structure, internal link targets, word count, CTA, and competitive gap analysis. Use when asked to "write a content brief", "brief this blog post", "plan this article", or "what should we cover for [keyword]".

Topic cluster architecture builder — takes a core topic and maps the full cluster with 1 pillar page, 6-10 supporting posts, internal linking map, keyword targets, and estimated monthly search volume per piece. Use when asked to "build a content cluster", "map our SEO cluster for [topic]", "create a topic cluster", or "what should our pillar page be about".

Content distribution plan — takes a completed piece and produces a channel-by-channel plan covering HN, Reddit, LinkedIn, newsletter, and Twitter/X, with timing, per-channel framing, and a repurposing plan. Use when asked to "distribute this post", "how do we promote this article", "write distribution copy for this piece", or "where should we share this".

Landing page and marketing copy — write hero section, problem/solution blocks, proof points, and CTAs. Use when asked to "write landing page copy", "write the homepage", "marketing copy for this feature", "product page copy", "write the hero section", or "write copy for [surface]".

|

Produce an actual launch plan with announcement copy, channel sequence, and day-1 checklist. Use when asked to "plan a launch", "GTM strategy", "how do we announce this", "launch plan for [feature]", "go-to-market", "write our Product Hunt post", or "how do we get people to notice this".

Messaging framework — produce a full headline, subheadline, proof points, and CTA hierarchy for use across all surfaces. Use when asked to "write our messaging", "messaging framework", "what should our headline say", "copy hierarchy", "tagline and messaging", or "how do we talk about the product".

Produce a complete positioning document using the Dunford framework — competitive alternatives, unique attributes, value, best-fit customer, market category, positioning statement, and tagline. Use when asked to "write our positioning", "define our value prop", "positioning statement", "what market are we in", "how do we position against X", "what's our tagline", or "write our messaging foundation".

Marketing and messaging reconnaissance — read existing landing pages, copy, positioning docs, and marketing materials to understand the current messaging state. Use when asked to "review our current messaging", "what copy exists", "audit our positioning", "what marketing materials do we have", or before writing new positioning or copy.

Cold outbound sequence builder — produces multi-touch email + LinkedIn sequences (5-7 touchpoints) personalized by persona type (technical buyer, economic buyer, champion). Use when asked to "write cold emails", "build an outbound sequence", "create prospecting emails", "write my LinkedIn outreach", or "design a cold email campaign".

B2B proposal generator — takes deal context (ICP, pain, pricing tier, timeline) and produces a complete proposal document with executive summary, problem statement, solution, pricing table, implementation timeline, ROI case, and next steps. Use when asked to "write a proposal", "draft our deck for this deal", "build a proposal for this customer", or "generate a proposal".

MEDDPICC-based deal qualification worksheet — guide Deal through structured qualification of any opportunity and produce a filled card + recommended next action. Use when asked to "qualify this deal", "run MEDDPICC on this opportunity", "should we pursue this", or "is this deal real".

Churn risk identification and intervention — scans health signals for at-risk accounts, classifies risk level (CRITICAL/HIGH/MEDIUM), and produces an intervention sequence per risk type. Use when asked to "find at-risk accounts", "who might churn", "build a churn prevention plan", "identify churn signals", or "rescue this account".

Quarterly Business Review template generator — takes account info (ARR tier, adoption metrics, goals) and produces a complete QBR deck outline and talking points. Use when asked to "prepare a QBR", "build a quarterly review", "write our QBR agenda", or "create QBR talking points for this account".

Customer segmentation model builder — tiers customers by ARR, health, and expansion potential; defines CS motion per tier; maps resource allocation. Use when asked to "segment our customers", "define our CS tiers", "how should we allocate CS resources", "build a customer segmentation model", or "who gets high-touch vs. digital".

|

Growth experiment design — structure a growth hypothesis, define metric, baseline, expected lift, and kill condition for a single experiment. Use when asked to "design a growth experiment", "test this growth idea", "experiment framework", "how do we test if this works", or "growth hypothesis".

|

PLG motion design — free tier definition, activation sequence, expansion trigger points, viral mechanic assessment. Given a product, output the PLG architecture and make the calls. Use when asked to "PLG strategy", "freemium model", "product-led growth plan", "self-serve motion", "how do we add a free tier", "upgrade triggers", or "viral loop design".

Growth state reconnaissance — scan existing onboarding flows, acquisition channels, conversion funnels, and growth experiment logs to understand current growth state. Use when asked to "what's our growth state", "audit the funnel", "what growth experiments have we run", "acquisition channel inventory", or before designing new growth experiments.

Retention diagnosis + intervention plan — analyze the retention curve, identify the primary drop-off point, and produce a specific intervention plan with expected impact. Use when asked to "improve retention", "why are users churning", "build a retention playbook", "reduce churn", "win-back campaign", or "users aren't coming back".

Plan and scope a project — discovery, challenge assumptions, present S/M/L options with token and cost estimates. Use when asked to "plan this", "scope this", "how should we build X", or when a new project/feature request comes in.

Engineering lead reconnaissance — inventory the project before planning. Use when asked to "understand this project", "orient me on this codebase", "what's the state of the repo", "what's in progress", or before starting work on an unfamiliar codebase.

Cross-cutting review of recent work — catches gaps between specialists. Use when asked to "review what we built", "check the work", "pre-launch review", or after completing a significant chunk of work.

CTO-level project status from git and codebase state. Use when asked "where are we", "project status", "what's done", or at the start of a work session.

System takeover — take ownership of an existing codebase or inherited system. Use when "we acquired this", "previous team left", "take over this system", "inherited this codebase".

Engineering lead — hand Apex any task and it routes internally. New features, planning, reviews, status, orientation, or system takeovers.

Write an Architecture Decision Record — document what was decided, why, what alternatives were considered, and what trade-offs were accepted. Use when asked to "write an ADR", "document this decision", or "why did we choose X".

Maintain per-repo and cross-repo changelogs — append structured entries after agent work. Use when asked to "log this change", "update changelog", "what changed", "change history".

Map the system architecture — read the codebase, identify services and connections, output a C4-level architecture map as Mermaid diagrams with component descriptions. Use when asked to "map the architecture", "system diagram", "how does this work", or "architecture overview".

Generate onboarding documentation — what this project does, how to set up locally, where things live, key decisions, how to deploy. Written for day-one engineers who know nothing. Use when asked for "onboarding docs", "new engineer guide", "how to get started", or "developer setup".

Generate a polished HTML presentation page and Obsidian Canvas for big releases — new products, takeovers, major migrations. Non-technical audience. Use when asked to "present this", "release announcement", "show what we built", or "stakeholder update".

Documentation reconnaissance for takeover — find all docs, assess accuracy, freshness, coverage, and discoverability, and identify critical knowledge gaps. Use when asked "what docs exist", "documentation assessment", or "knowledge gaps".

Render agent findings as a styled HTML report in the browser. Use when asked for "full report", "detailed report", "show in browser", or when CLI output exceeds the 40-line budget.

Knowledge engineer — architecture docs, ADRs, diagrams, changelogs, onboarding, and reports.

Build and manage open source community — Discord/Slack structure, contributor onboarding, ambassador program, community flywheel design, and GitHub community health. Use when asked to "build a community", "grow our Discord", "improve contributor experience", or "design a developer ambassador program".

Design and execute a launch plan — Product Hunt, HN Show HN, newsletter coordination, social posts, and community launch moment. Use when asked to "launch [feature/product]", "plan a launch", "help us do a Product Hunt launch", or "coordinate the announcement".

Write media pitches and press releases — journalist outreach emails, podcast pitch scripts, newsletter sponsor pitches, and press release copy. Use when asked to "pitch journalists", "write a press release", "reach out to podcasts", or "get media coverage".

PR and community reconnaissance — audit current press coverage, social presence, community health, and competitor PR. Use when asked to "audit our PR", "what's our community state", "how do we compare in press", or before planning a launch or community initiative.

Social media strategy and post drafting — HN posts, Twitter/X threads, LinkedIn posts, Reddit comments, and developer community content. Use when asked to "write a HN post", "draft social posts", "help us post on Twitter", or "create a social launch plan".

PR & Community engineer — press pitches, social media, open source community, DevRel, and coordinated launch moments.

Evaluate model performance — check for accuracy drops, data drift, and error patterns. Use when asked about "model accuracy dropped", "evaluate the model", "check for drift", or "model performance".

Design and implement an AI feature integration — model selection, architecture pattern, system prompt, data flow, error handling, cost estimate. Use when asked to "add AI to this", "LLM integration", "add Claude/GPT", or "AI-powered feature".

Build an ML pipeline — from data to trained model to serving endpoint. Use when asked to "build ML model", "train a model", "prediction pipeline", "classification", or "regression".

Build a production-ready prompt package — system prompt, few-shot examples, output format, edge case handling, eval criteria. Use when asked to "prompt engineering", "build a prompt", "write a system prompt", or "improve this prompt".

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

Review UI code for Web Interface Guidelines compliance. Use when asked to "review my UI", "check accessibility", "audit design", "review UX", or "check my site against best practices".

Use this skill when you need to control a Chrome browser via CDP (Chrome DevTools Protocol) to reuse existing login sessions. Covers: launching Chrome in debug mode, opening URLs, waiting for page load, evaluating JavaScript, taking snapshots, and extracting auth tokens. Trigger phrases: browser automation, CDP, agent-browser, 浏览器操作, 操作浏览器, Chrome CDP, 复用登录态, extract token from browser.

|

|

|

|

|

|

|

|

|

|

|

|

当用户明确要求"核查/优化综述 `{主题}_review.tex` 的正文引用"或"运行 check-review-alignment"时使用。通过宿主 AI 的语义理解逐条核查引用是否与文献内容吻合，只在发现致命性引用错误时对"包含引用的句子"做最小化改写，并复用 `systematic-literature-review` 的渲染脚本输出 PDF/Word。核心原则：不为了改而改，无法确定是否为致命性错误时保留原样并在报告中警告。⚠️ 不适用：用户只是想生成系统综述正文（应使用 systematic-literature-review）；用户只是想新增/核对 BibTeX 条目（应使用专门的 bib 管理流程）。

当用户明确要求"填充示例内容""生成示例""补充 LaTeX 示例"时使用。AI 增强版 LaTeX 示例智能生成器，实现 AI 与硬编码的有机融合：AI 做"语义理解"（分析章节主题、推理资源相关性、生成连贯叙述），硬编码做"结构保护"（格式验证、哈希校验、访问控制）。

当用户明确要求"从文件/图片/网页/描述中提取综述主题"或"生成主题+关键词+核心问题结构化输出"时使用。支持文件（PDF/Word/Markdown/Tex）、文件夹、图片、自然语言描述、网页 URL 等多种输入源，自动识别输入类型并提取内容，生成可直接用于 systematic-literature-review 及其他文献综述技能的结构化输出。