guanyang/antigravity-skills
82 skills7.9k installs77.2k starsGitHub
Install
npx skills add https://github.com/guanyang/antigravity-skillsSkills in this repo
1Filesystem Contextfilesystem-context is an antigravity-skills agent skill (version 1.2.0) that treats the filesystem as the primary overflow layer for coding-agent context engineering. It documents six patterns—scratch-pad offloading above a ~2000-token threshold, YAML plan persistence, sub-agent workspace directories, dynamic SKILL.md loading, terminal log files, and preference self-modification with validation guards. The guide diagnoses four context failure modes (missing, under-retrieved, over-retrieved, buried) and prescribes grep, glob, and line-range reads for selective retrieval. Developers reach for filesystem-context when tool outputs bloat prompts, multi-turn tasks need durable state, or sub-agents must hand off findings without message-chain summarization loss. Eleven guidelines and eight gotchas cover retention policies, race conditions, and path hygiene for production agent workflows.406installs2Planning With Filesplanning-with-files is a Manus-style agent skill in guanyang/antigravity-skills (metadata version 3.1.3) that writes planning state to three canonical markdown files instead of volatile context windows. Invoked via /planning-with-files or @planning-with-files, it creates task_plan.md for phased goals, findings.md for research discoveries, and progress.md for session logs, with lifecycle hooks on UserPromptSubmit, PreToolUse, PostToolUse, Stop, and PreCompact events. Developers reach for planning-with-files when organizing multi-step projects, research tasks, or any agent work requiring five or more tool calls where /clear, crashes, or context compaction would otherwise lose progress. The skill auto-restores context by re-reading planning files at session start and supports session-catchup scripts to reconcile unsynced state after interruptions.340installs3Ui Ux Pro MaxAntigravity ui-ux-pro-max skill guides agents to ship polished SaaS, mobile, and content interfaces—covering layout grids, typography, color, spacing, component variants, hover/focus states, and accessible interaction patterns so outputs feel intentional rather than generic boilerplate.226installs4Memory Systemsmemory-systems guides implementation of agent memory stacks: what to store, how to retrieve and summarize context, how to bound token use, and how to keep multi-turn assistants coherent across sessions and tool calls.208installs5Frontend Designfrontend-design helps teams produce cohesive, production-ready web interfaces by defining layout systems, visual hierarchy, branding-aligned tokens, and responsive component patterns during frontend build work.170installs6Context CompressionAntigravity skill focused on compressing conversational and codebase context for LLM agents. Reduces token usage while retaining critical instructions, decisions, and task state for reliable long-horizon automation.165installs7Pptxpptx enables agents to read, create, and modify PowerPoint files: slide layouts, charts, speaker notes, and branded masters for pitches, launch decks, training, and stakeholder presentations ready for distribution.154installs8XlsxXlsx skill handles spreadsheet-centric tasks end-to-end: parse messy tables, add computed columns and charts, apply formatting, and save polished Excel or CSV outputs for finance, ops, and content teams who require auditable workbook deliverables.138installs9BrainstormingGuides Claude through disciplined brainstorming sessions that turn vague goals into ranked options, explicit assumptions, and decision-ready next steps before any prototype or implementation work begins in antigravity-skills workflows.127installs10Canvas DesignCanvas-design helps agents plan and implement canvas-driven visuals—layouts, illustrations, charts, and interactive graphics—in web frontends with correct sizing, layering, and performance patterns for polished UI moments.127installs11Systematic DebuggingImplements a rigorous four-phase debugging methodology in antigravity-skills that forces reproduction, narrowing, hypothesis testing, and post-fix verification so agents and developers resolve bugs with evidence instead of speculative patches.127installs12Theme FactoryTheme-factory skill generates end-to-end visual themes—colors, fonts, spacing, and component styling rules—for web UIs so Claude Code can apply a single cohesive design language across SaaS dashboards, marketing pages, and extensions.125installs13Web Artifacts BuilderAntigravity web-artifacts-builder skill helps agents construct self-contained web artifacts—HTML, CSS, and JS demos, widgets, or micro-sites—for preview, embedding, or review without bootstrapping a full application, accelerating frontend iteration on SaaS, content, and extension surfaces.125installs14DocxDocx guides agents to read, create, and edit Microsoft Word files with proper styles, tables, headings, and templates—useful for reports, proposals, and compliance documents that must remain editable in Office workflows.124installs15Using SuperpowersMeta-skill for Claude Code superpowers: how to chain skills, spawn subagents, wire MCP tools, and apply verification habits so builders consistently get multi-step refactors, reviews, and integrations done with less manual orchestration.123installs16Multi Agent PatternsDocuments proven multi-agent orchestration patterns for antigravity-skills so teams can pick the right topology, define agent roles, manage shared context safely, and avoid common coordination failures in production agent builds.121installs17Brand GuidelinesBrand-guidelines teaches agents how to apply a product's visual and verbal identity—color, typography, logo rules, and tone—so generated interfaces, decks, and copy stay consistent from first mock through launch.120installs18Algorithmic ArtAntigravity skill for creating algorithmic and generative art through code rather than static design files. Helps developers produce unique visuals, interactive sketches, and procedural graphics for web, games, and creative demos.118installs19PdfPDF skill from guanyang/antigravity-skills equips agents to read, extract, manipulate, and generate PDF documents during product development. It targets workflows where specs, invoices, contracts, or user uploads arrive as PDFs and must become structured data, filled forms, or exported reports inside SaaS and API products.118installs20Remotion Best Practicesremotion-best-practices guides building data-driven videos in React with Remotion: composition architecture, frame-accurate animation, asset handling, rendering strategies, and production patterns for explainers and automated media.117installs21Verification Before CompletionAntigravity verification-before-completion skill requires agents to verify work before declaring done—running tests, reproducing bugs, validating acceptance criteria, and confirming integrations behave—reducing broken handoffs across SaaS, API, and agent workflows.117installs22Notebooklmnotebooklm equips agents to run NotebookLM-like research flows: upload or reference source corpora, ask grounded questions, produce cited summaries, and turn messy early reading into actionable ideation notes.116installs23Skill CreatorGuides creation of installable Claude Code skills: SKILL.md layout, invocation triggers, optional scripts/references, and disclosure tiers so agents consistently run bespoke build, review, or ops workflows in antigravity-skills style repos.116installs24Writing PlansWriting-plans from guanyang/antigravity-skills guides creation of clear implementation plans with milestones, dependencies, risks, and acceptance criteria during scope validation. It helps PM-style workflows for SaaS, API, and agent projects so teams and coding agents share one executable blueprint before development spend accumulates.116installs25Executing PlansProvides a structured antigravity-skills workflow for executing written implementation plans step by step, reporting status at checkpoints, and adjusting order or scope without losing alignment to the original design intent.115installs26Mcp BuilderMcp-builder walks through creating Model Context Protocol servers—tool schemas, transports, authentication, and packaging—so agents can call Stripe, databases, or internal APIs through well-typed, testable integration surfaces.115installs27Context OptimizationContext-optimization teaches strategies to trim, rank, and compress agent context—summaries, selective retrieval, and deduplicated tool results—so long coding sessions stay within token limits while preserving task-critical information.114installs28Finishing A Development BranchAntigravity skill for completing a development branch and preparing it for merge. Covers final review steps, PR structuring, conflict handling, and cleanup so features ship through git workflows without loose ends.114installs29Subagent Driven DevelopmentSubagent-driven-development from guanyang/antigravity-skills teaches orchestration patterns for breaking implementation into delegated subagent runs with clear boundaries, parallel execution, and structured handoffs. It fits agentic coding workflows building SaaS, CLIs, and autonomous tools where one agent cannot safely own an entire feature surface alone.114installs30Webapp TestingWebapp-testing skill directs Claude Code through browser-based QA: start dev servers, navigate flows, assert UI state, capture failures, and iterate fixes so SaaS and extension frontends ship with regression coverage on auth, forms, and navigation paths.114installs31Web Design GuidelinesWeb-design-guidelines from guanyang/antigravity-skills provides practical UI and UX standards for web prototypes covering typography, spacing, color, hierarchy, and accessibility. It supports validating SaaS, ecommerce, and content experiences early so mocks and landing experiments look cohesive before frontend implementation begins.114installs32Obsidian MarkdownObsidian-markdown teaches agents to write vault-ready notes with wikilinks, callouts, tags, embeds, and YAML frontmatter so generated documentation imports cleanly into Obsidian knowledge bases and backlink graphs.113installs33Receiving Code Reviewreceiving-code-review teaches developers and agents how to process pull request feedback professionally: categorize comments, plan fixes, avoid argumentative loops, validate with tests, and return clear responses that speed merge.113installs34Test Driven DevelopmentTest-driven-development from guanyang/antigravity-skills encodes red-green-refactor discipline for agents building SaaS backends, APIs, and CLIs. It prioritizes failing tests first, minimal passing implementation, and refactor steps so new behavior ships with regression coverage instead of retrofitted test suites after features are written.113installs35Writing SkillsAntigravity writing-skills meta-skill teaches authoring Claude Code skills—defining triggers, structured workflows, bundled references, and progressive disclosure—so teams turn expertise into reusable SKILL.md packages agents load consistently across repositories.113installs36Vercel Composition PatternsVercel-composition-patterns from guanyang/antigravity-skills documents composable frontend architecture for Next.js and React on Vercel, covering server-client splits, layout composition, and reusable UI structure. It helps teams building SaaS and content sites avoid monolithic pages and adopt patterns that scale routes, shared shells, and deployment on Vercel.112installs37Doc CoauthoringAntigravity skill for collaborative document writing with an AI partner. Guides iterative drafting and editing of specs, RFCs, READMEs, and internal docs so engineering teams ship clearer documentation faster.110installs38Json Canvasjson-canvas teaches agents to generate and maintain JSON Canvas documents: node graphs, connectors, embedded media, and cross-linked visual maps suitable for Obsidian and other canvas-compatible knowledge tools.110installs39Requesting Code ReviewTrains antigravity-skills users to package code changes for human or agent reviewers with concise context, highlighted risk areas, reproduction steps, and targeted questions that shorten review cycles and improve review quality.109installs40Using Git WorktreesTeaches git worktree workflows for parallel development: spawn branch-specific directories, coordinate multiple agent tasks, and manage lifecycle cleanup so teams ship stacked features without constant context switching on a single working tree.109installs41Obsidian Basesobsidian-bases helps agents set up and refine Obsidian Bases: property schemas, filtered table views, formulas, and rollups that operationalize notes into trackable backlogs, specs, and status dashboards over time.108installs42Dispatching Parallel AgentsTeaches antigravity-skills users how to decompose work into parallel agent dispatches with explicit boundaries, handoff rules, and merge expectations so Claude can execute independent chunks simultaneously without cross-agent conflicts.107installs43Internal CommsStandardizes how Claude writes internal communications in antigravity-skills: clear headlines, relevant context, explicit asks, and consistent formatting for status, incidents, and change announcements across engineering and product teams.107installs44Vercel React Best Practicesvercel-react-best-practices is a Claude Code skill for frontend development. It helps solo builders move faster with AI-assisted development.107installs45Context FundamentalsContext-fundamentals explains how LLM context windows, message roles, and token budgets interact so builders design agent prompts, tool outputs, and skill loading with predictable limits and fewer truncation surprises.106installs46Supabase Postgres Best Practicessupabase-postgres-best-practices is a Claude Code skill for databases. It helps solo builders move faster with AI-assisted development.105installs47Vercel React Native Skillsvercel-react-native-skills is a Claude Code skill for frontend development. It helps solo builders move faster with AI-assisted development.104installs48Advanced EvaluationA skill covering production-grade LLM-output evaluation, choosing between direct scoring and pairwise comparison and mitigating known judge biases. A developer uses it to build eval pipelines, compare model outputs, or design scoring rubrics.101installs49Project DevelopmentA methodology for scoping LLM projects: assess task-model fit, prototype manually, then build discrete idempotent pipelines with cost control. A developer uses it when starting an LLM-powered application or batch-processing pipeline.101installs50Defuddledefuddle is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.98installs51Bdi Mental StatesTransforms external RDF context into agent mental states (beliefs, desires, intentions) using formal BDI ontology patterns and a triples-to-beliefs-to-triples pipeline. A developer uses it when building cognitive multi-agent systems that need traceable, explainable reasoning.96installs52Hosted AgentsCovers infrastructure for running coding agents in remote sandboxed VMs, from image pre-building and warm pools to self-spawning agents and multi-client interfaces. A developer uses it when building background agents that run independently of local machines.96installs53Obsidian Cliobsidian-cli is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.96installs54Tool DesignTreats each tool as a contract an agent must infer from its description alone, favoring consolidation and architectural reduction over many narrow tools. A developer uses it when creating or debugging tools for agent or MCP systems.96installs55Context DegradationProvides detection signals and mitigation strategies for the five ways agent context degrades, treating degradation as a measurable engineering problem. A developer uses it when agent quality drops during long conversations or large-context tasks.94installs56EvaluationExplains how to evaluate non-deterministic agents by scoring outcomes across weighted dimensions rather than single metrics or fixed paths. A developer uses it when building quality gates, catching regressions, or comparing agent configurations.93installs57Claude Apiclaude-api is a Claude Code skill for backend & apis. It helps solo builders move faster with AI-assisted development.75installs58Latent BriefingDescribes latent briefing, a representation-level memory-sharing technique that compacts the orchestrator trajectory in the worker's KV cache with task-guided queries. A developer uses it to cut worker tokens in hierarchical multi-agent systems where they control the inference runtime.57installs59Baoyu ImagineAn image-generation skill wrapping official APIs (OpenAI GPT Image 2, Azure, Google, OpenRouter, DashScope, GLM-Image, MiniMax, Jimeng, Seedream, Replicate), sequential by default with batch parallel support. A developer uses it to create or draw images after a blocking preferences setup step.42installs60Baoyu Translatebaoyu-translate is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.36installs61Baoyu Wechat Summarybaoyu-wechat-summary is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.33installs62Baoyu Image Cardsbaoyu-image-cards is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.29installs63Baoyu Comicbaoyu-comic is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.26installs64Baoyu Cover Imagebaoyu-cover-image is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.26installs65Baoyu Diagrambaoyu-diagram is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.26installs66Baoyu Infographicbaoyu-infographic is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.26installs67Baoyu Compress Imagebaoyu-compress-image is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs68Baoyu Danger Gemini Webbaoyu-danger-gemini-web is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs69Baoyu Format Markdownbaoyu-format-markdown is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs70Baoyu Image GenAn image-generation skill wrapping many official provider APIs, supporting text-to-image, reference images, aspect ratios, and batch generation from prompt files. A developer uses it to create images sequentially or in parallel batches after a blocking preferences setup.25installs71Baoyu Markdown To Htmlbaoyu-markdown-to-html is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs72Baoyu Post To Wechatbaoyu-post-to-wechat is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs73Baoyu Post To Weibobaoyu-post-to-weibo is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs74Baoyu Post To Xbaoyu-post-to-x is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs75Baoyu Slide Deckbaoyu-slide-deck is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs76Baoyu Xhs Imagesbaoyu-xhs-images is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs77Baoyu Youtube Transcriptbaoyu-youtube-transcript is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.25installs78Replace With Skill Name Using HyphensAn empty SKILL.md template with placeholder headings for overview, triggers, core pattern, workflow, and verification checklists. A developer copies it as a starting point when writing a new skill.25installs79Baoyu Article Illustratorbaoyu-article-illustrator is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.24installs80Baoyu Danger X To Markdownbaoyu-danger-x-to-markdown is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.24installs81Baoyu Url To Markdownbaoyu-url-to-markdown is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.24installs82Baoyu Electron Extractbaoyu-electron-extract is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.20installs