Claude Skills
The open format is called Agent Skills and works in Claude Code, Codex, Cursor and other agents — most people know it as Claude Skills.
Every Agent Skill we could find on GitHub, deduplicated by content. 79 600 files from 1 763 authors, of which 61 947 are unique — the rest is the same skill repackaged into someone else's repository. For each one: what it weighs in tokens, whether it ships runnable scripts, and which MCP servers it needs.
Guide user through errors they see. Translate error messages to human language.
After fixing a bug, explain what went wrong in language the user understands. No jargon for beginners.
Step 2 of debugging: reason about root cause based on observed symptoms.
Step 3 of debugging: apply minimal fix. One-liner ideal. Verify nothing else broke.
Final go-live checklist. Is it accessible? SSL working? Mobile friendly? Show user their live URL.
Detect recurring patterns in user's project. Apply automatically next time.
Full project review from beginning. Check every file. Plain language report.
Learn user's preferences over time. Code style, naming, structure. Save to global prefs.
Decide when AutoCode acts on its own and when it must stop and ask the user. Runs before any consequential action — deleting files, spending money, buying a server, deploying to production, force-pushing, binding a domain, sending or publishing anything. Ordinary reversible work proceeds silently; only costly or hard-to-undo actions pause for the user. Powered by I-Lang v5.0 vector judgment.
Persistent memory across sessions. Save project state and user preferences. Never save secrets.
Record mistakes. Check before similar builds. Avoid repeating silently.
Help user work from multiple devices. Sync project via git.
Break complex tasks into 5-15 ordered steps with time estimates. Dependency order first.
Track project milestones. Auto-detect when a significant checkpoint is reached.
Give realistic time and cost estimates for each step. Explain in human terms.
Identify risks before building. Flag third-party dependencies, API limits, and cost traps.
Report progress after each feature. Percentage, what just completed, what comes next.
Decide what to build first. Core function before polish. Revenue before aesthetics.
Optimize for speed and cost. Pick lightweight solutions. Flag expensive operations.
Create a visual roadmap for large projects. Phases, milestones, timeline.
Lock confirmed requirements. Don't change them without user approval.
Run and test directly on the server. No local dev environment needed. What you build is what goes live.
Save checkpoints before risky changes. Rollback if things break.
Detect user's intent from their message and activate the right workflow silently.
Auto-apply security basics and block deploys that would leak secrets. Never ask the user about security choices — just do it, and run a real secret scan before going live.
Detect user's technical level from first messages. Adjust all output language accordingly.
Operate ArvanCloud APIs, DNS records, Cloud Server/IaaS, Object Storage, VOD/Live, CDN/Security, and Let's Encrypt DNS-01 certificates via acme.sh dns_arvan.
Draft an Instagram caption with the hook in the first 125 characters, a skimmable body, and one clear CTA, using a 2026 hook formula chosen by goal (saves, shares, comments, follows). Runs the humanizer pass and, on approval, publishes with the image or video you supply via the Publora media flow. Use to caption a single image or a Reel. Not for slide-by-slide carousels (use ig-carousel-planner) or auditing a draft (use ig-humanizer --mode audit).
Plan an Instagram carousel slide by slide, up to 10 slides, with a hook slide that opens a loop, value slides that front-load the payoff, and a payoff slide that earns the save and follow. Picks a 2026 carousel formula (listicle, before/after, myth-buster, framework) by goal (saves, shares, follows), drafts each slide's text plus the caption, then publishes the images you supply via Publora. Use to structure a multi-slide carousel. Not for single-image captions (use ig-caption-writer).
Generate a weekly Instagram content plan from a theme, audience, and content pillars. Produces a per-day mix of Reels, carousels, single images, and stories, each with a hook formula, angle, posting time, and primary goal, plus a weekly saves-and-shares goal (the 2026 lever), a story cadence, and a balance check across saves, shares, comments, and follows. Use to plan a week instead of ad-hoc posting. Not for drafting one caption (use ig-caption-writer) or one carousel (use ig-carousel-planner).
Build a sized Instagram hashtag set for a post using the 2026 reality that 3 to 5 well-chosen tags beat 30 random ones. Sizes each tag as niche (under 50k posts), mid (50k to 500k), or broad (500k+), then assembles a 2-3 niche, 1-2 mid, 0-1 broad mix that a small or mid account can actually rank in, matched to the post content and rotated to avoid spam patterns. Use to choose hashtags for a caption, carousel, or Reel. Not for writing the caption itself (use ig-caption-writer).
Read your Instagram niche and profile from real data via Apify, no login. Scan a hashtag for the posts traveling now (likes, comments, owner) to see the format and hook that works. Pull profile stats for any handle, yours or a competitor's: followers, posts, bio, category. Instagram hides who liked or commented on other accounts, so this is discovery plus profiles, not engagers. Triggers on \"what works in my niche\", \"scan the hashtag\", \"competitor stats\". Not for writing captions (use ig-caption-writer).
Plan, draft, audit, and publish content for Instagram. Use when the user wants to write a caption with a first-125-char hook, plan a slide-by-slide carousel, reverse-engineer the hook from a viral Reel or carousel, size hashtags the 2026 way, remove AI tells from a caption, or plan a week of Reels and carousels. Instagram requires media on every post, so the skills write the caption and the user supplies the image or video; posts publish via the Publora API draft, upload, schedule flow.
Remove the AI tells readers react to in an Instagram caption or carousel slide: 2026 vocabulary by density, reveal bridges, staccato stacks, stacked triads, performed sincerity, emoji storms; caps em dashes. Includes --mode audit (first-125 hook, length, hashtags, emoji, CTA, media) and --mode profile. Not for beating AI detectors (no edit reliably does). Not for writing from scratch (use ig-caption-writer or ig-carousel-planner). Keywords: humanize, de-AI caption, audit before posting.
Reverse-engineer the hook from a viral Instagram Reel or carousel, given a URL or a pasted caption. Identifies which of the 10 canonical 2026 formulas it uses (number-first, contrarian, relatable, confession, listicle, before/after, myth-buster, framework, pattern-interrupt Reel, how-I Reel), explains why it worked, names the primary goal, and returns a blank template mapped to your topic. Use to learn from a post you admire. Not for writing your own (use ig-caption-writer or ig-carousel-planner).
Audit and rewrite an Instagram profile end-to-end for 2026: profile photo, searchable NAME field weighted with a keyword, @handle, bio (150 chars: value plus topic plus proof), goal-matched link, category label, story highlights, the first-9 grid, and up to 3 pinned posts. Triggers on \"review my Instagram profile\", \"fix my bio\", \"optimize highlights\", \"profile audit\". The whole follow decision happens on the profile header. Not for writing captions (use ig-caption-writer).
Repurpose existing content into a native Instagram post. Take a LinkedIn, blog, YouTube script, or X tweet or thread and rebuild it for Instagram: a long piece becomes a carousel or a caption, re-hooked before the 125-char fold, off-platform artifacts stripped (X @-handles, link in bio), published via Publora on approval. Use to adapt content across platforms into Instagram. Not for writing from scratch (use ig-caption-writer or ig-carousel-planner), not for auditing a draft (ig-humanizer --mode audit).
Build interactive, click-through architecture diagrams for software systems — a self-contained single HTML file with animated step-by-step flows, mode toggles (dev/prod, offline/online, v1/v2), dark/light theme, and a side panel with payload details, plus a companion markdown description. Use when the user wants to visualize or design a system: architecture diagram, service map, data flow, RAG/agentic flow, microservices topology, integration diagram, CI/CD or data/ETL pipeline, multi-agent system, or onboarding diagram — or to design a new service's flows before building. Natural-language triggers: 'I'm building X, what should the flow look like', 'pokaż jak działa system', 'diagram architektury', 'wizualizacja systemu', 'klikany diagram'. Do NOT use for static diagrams that belong inline (Mermaid/PlantUML), slide decks, or printable/PDF handouts — this produces interactive HTML for browser consumption.
Generate synthetic training data using NVIDIA Cosmos world foundation models for SDG pipelines
Monitor robot fleet telemetry via Azure IoT Operations, drift detection, Grafana dashboards, and Fabric analytics
Generate, transfer, and consume environment-specific Azure, AKS, OSMO, ACR, and Azure ML deployment bundles. Use when: discovering deployed environment details; creating OSMO image manifests or platform values; preparing a HiL host; uploading or downloading deployment files through Azure Key Vault; or deploying with generated environment configuration.
Deploy and manage Azure infrastructure for the Physical AI Toolchain including Terraform IaC, Kubernetes setup, GPU configuration, and network topology
Deploy trained robot policies to edge fleets via FluxCD GitOps, image automation, and deployment gating
Start and interact with the Dataset Analysis Tool (dataviewer) for browsing, annotating, and exporting robotic training episodes
Submit, monitor, analyze, and evaluate LeRobot imitation learning training jobs on OSMO with Azure ML MLflow integration and inference evaluation - Brought to you by microsoft/physical-ai-toolchain
Use when users request Android competitor research, systematic feature or interface mapping, evidence screenshots, interaction coverage, comparison-ready product dossiers, or Android system-boundary verification without APK decompilation.
Design, redesign, review, or implement user interfaces using Google Material Design 3 (M3), Material You, and optional Material 3 Expressive principles. Use for UI/UX screens, app flows, design systems, component selection, theming, responsive/adaptive layouts, accessibility audits, Figma-ready specifications, or developer handoff where Material Design 3 is required.
以“成事而不妄为”为入口处理通用任务。用户点名 odai，或任务模糊、复杂、高风险、前提可疑、易漏项或需跨能力协作时使用；简单任务直做。
日报、周报、工作总结、状态更新：基于可核证事实形成完整汇报。用户要求汇报一段时期的工作、沿用既有汇报模板、整理进展与风险，或 odai 判断专业汇报工艺会改善结果时使用。
Shape output for a reader with ADHD, then write each sentence in controlled English: action first, one word one meaning, active voice, simple tenses, state restated every turn. Invoke with /attention-control; stays on until "stop attention control".
> 论文写作、修改、压缩和实验组织全流程中阻止【防御性写作】。把论文视为一场学术发布会， 而不是项目总结、实验日志或自我审查报告：识别这项工作最值得发表的价值，围绕它建立 最有利、最完整、最有说服力的叙事。不平均展示、不主动示弱、不写实验流水账、不替审稿人 攻击自己。 触发词：论文润色、论文修改、写摘要、写引言、写结论、压缩论文、rebuttal、回复审稿意见、 组织实验、论文AI味、防御性写作、心虚。
> Stops defensive writing across the entire paper lifecycle — writing, revising, cutting, and organizing experiments. Treats the paper as a press conference, publishable strength of the work and build the most favorable, complete, and persuasive narrative around it. Never present everything evenly, never volunteer weakness, never write an experiment diary, never attack the paper on the reviewer's behalf. write conclusion, shorten paper, rebuttal, respond to reviewers, organize experiments, defensive writing, AI-flavored academic writing.
Act on LinkedIn through an account you own, over MCP or REST, using gtm-api.com. Use when a task involves LinkedIn search, connection requests, messages and InMail, inbox reading, profile or company enrichment, post engagement, or checking what a LinkedIn account is still allowed to do today. Covers connecting the server, the three meta-tools and their discovery order, the preview-and-confirm gate on outward actions, and how to read a limit rejection instead of retrying it.
Back an app up into one zip that lands in the user's own Downloads folder with no storage permission, keeping only the newest N archives. Covers checkpointing the database's write-ahead log before the file is copied, assembling the archive in cache first, inserting through the system media store, and pruning old archives with a query over that same collection. Android only. Use when a restored backup is missing the most recent writes, when the backup file is invisible to the user's file manager, or when old backups accumulate forever.
Audit every native dependency for a slice on a CPU architecture before promising that target in a multiplatform desktop build — one missing native takes the whole target down at first use rather than at build time, so make the audit a repeatable command over the resolved artifacts and re-run it on every dependency bump; reach for it when deciding whether to add an ARM64 target, or when a build that packaged and installed cleanly dies the first time it touches the database, the renderer or the media layer.
A remote write can answer "ok" and still have discarded what you sent, saying so only in a secondary field riding along with the success. Model accepted-but-discarded as its own outcome, read that field on every write, and log a discard loudly. Reach for it when a submission reports success on every call and the data never appears on the other side.
Draw a linear gradient at an arbitrary angle across a Compose box so both endpoints land exactly on the box edge — the per-quadrant endpoint formula from the requested angle, why rotating the diagonal overshoots and why clamping to the nearest edge distorts the angle, and the degenerate cases that collapse the ramp to nothing. Use when a tilted gradient looks washed out or cut off near the corners, when the visible angle does not match the angle you asked for, or when the same gradient looks different on a wide box than on a tall one.
A design rule with legitimate exceptions survives only if every exception carries its reason at the call site and the rule itself is greppable — otherwise nothing distinguishes an exception from a violation and the rule silently rots. Covers where the rule statement goes, where the reasons go, scoping the audit to the code the rule actually governs, and the limits of a comment-based check. Use when a stated convention is drifting, when reviewers cannot tell deliberate from careless, or before writing a rule into a file header and assuming it will hold.
A rotating sweep-gradient ring around a card, chip or button, built from a clipped box, a full-size gradient and an opaque inset surface parked on its middle — where the modifier order is the whole mechanism and no blend mode is involved — plus why the SrcIn-inside-an-offscreen-layer version of this cannot draw a ring, and what that layer costs. Use when the gradient covers the whole surface instead of the border, when the ring is invisible, when it bleeds over whatever is behind the widget, or when a row of them makes scrolling expensive.
Claude Skills — questions
Answers built from the skills we actually parsed.