Skills
96,770 packaged capabilities and playbooks for agents. Install any of them with one command.
All
The front door of cooked. Takes the user's video and a few multiple-choice answers (platform, creators to copy and what to take from each, audience, brand kit, own photos, names, what to avoid, captions, sound), saves them once to a profile, then runs the whole edit in order (setup if a tool is missing, creator-teardown for a new creator, cut, style-edit) so the user only drops a video, answers, approves the cut and approves the stills. Use when the user says "edit my video", "edit this like @creator", "make my video look like @creator", "make this look like <creator>'s videos", "I want my videos edited", "start", "new video", drops a video file or pastes any link (video, Drive or Dropbox share, creator profile, website, podcast) with no other instruction, or asks how to use the editor. Not for one step named on its own (only cutting is cut, only analysing a creator is creator-teardown, a correction is taste, styling an existing cut is style-edit).
Turns an approved cut into the finished video in a creator's measured style. Takes cut.mp4 from the cut skill plus a style.json (from start or creator-teardown), plans captions and zooms to match the creator, adds visuals timed to the spoken word (the user's own images, real screenshots, posts and logos of what is named, the line marked on the real page; no type cards, stock icons or emoji), shows the stills on the review page for approval, then renders 9:16 or 16:9 with Remotion on the laptop, Modal, GitHub Actions or the user's AWS Lambda, and can export the edit to Final Cut, Premiere, Resolve or CapCut. Use when a cut already exists and the user says "style it", "now add the style", "add captions and zooms", "add my images to the video", "render the edit", "change the captions", "export to Final Cut" or "send it to Premiere". Not for a raw take or a first request to edit a video like a creator (that is start), not for cutting mistakes out (cut), and not for measuring a creator (creator-teardown).
Run an extremely strict maintainability review focused on abstraction quality, oversized files, and growth of spaghetti conditions. Use for a thermo-nuclear code quality review, thermonuclear review, deep code quality audit, or especially harsh maintainability review.
Use when writing or reviewing working text: a repository page, a code comment, a description or help text, a PR, an issue, a CI comment, a Notion page, or a diagram that a colleague will scan to get work done, or when a change moves code across the layers an architecture page declares. Not for a paper, an essay, or any text read start to finish.
Act as the Light workflow advisor, navigator, and router. Understand what the user is trying to do, inspect the actual project/host evidence and Skill catalog, reason about the best next Light Skill, explain why, wait for user approval, then honor the accepted Skill's invocation policy with host-aware transition. Also answer collection-navigation and standalone routing requests. Use only when the user explicitly invokes $ask-light; never execute before consent
Decompile a Blitz3D game to readable Blitz source with B3DDecomp and port what it does to another engine. Covers SCP - Containment Breach-era and other .bb games, B3DDecomp, Blitz3D runtime behaviour (fixed tick, Animate, CurveValue, Euler angles, quaternions, Rnd), animated .b3d models (viewmodels with bones and keyframes) into Roblox, and where to stop (builds protected against decompiling, games and assets the user has no rights to). Use when a Blitz3D game's code or animated models are needed.
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, define verifiable success criteria, keep a WORKLOG.md so work can be continued across sessions, and write research to research/ and plans for review to docs/.
Query KIRHA — a dynamic catalogue of authoritative data providers accessible from the command line. USE THIS SKILL whenever a question would be answered better by an authoritative source than by model memory or a generic web search: facts that change over time, provenance or citability, official records or canonical registries, claims to verify or fact-check, current documentation, or whenever accuracy beats guesswork. KIRHA's catalogue grows continuously and coverage is discovered at runtime, so do not assume in advance which domain applies. Prefer this skill over WebSearch when accuracy, freshness, or source-traceability matters. KIRHA returns structured facts from named providers, not scraped page summaries. Trigger EVEN IF the user does not mention KIRHA or "search" — if an authoritative source would answer better, this is the skill
Turn conversations, transcripts, project notes, folder workflows, documentation, or other sources into reusable agent skills and Skill Creator compatible instructions. Use when a user asks to create, update, distill, or "learn" a workflow as a skill; preserve repeated operating methods for future AI agents; encode corrections/failure modes as durable procedure; or transform source material into portable skill guidance for coding agents, research agents, automation agents, or custom agent frameworks
Use to build, manage, and improve the context that makes Hex's AI agents (Threads, the Notebook Agent, the Modeling Agent) accurate — endorsements & exclusions, warehouse descriptions, workspace context & guides, and semantic models. Covers standing context up from scratch, measuring it with a dedicated eval engine (authoring and running eval suites, sourcing cases from real usage, gating changes in a preview), reviewing Context Studio Suggestions, and publishing via GitHub sync or the Hex CLI. Triggers on: “context management”, “manage my Hex context”, “context strategy”, “context engineering”, “set up Threads context”, “workspace guide”, “warehouse descriptions”, “endorse tables”, “semantic model”, “my agent gave the wrong answer”, “audit my Hex context”, “improve agent accuracy”, “eval my context”, “write Hex evals”, “eval rubric”, “baseline vs hill-climbing”, “convert my evals”, “test context changes”, “hex context preview”, “context suggestions”, “Hex Context Studio”. Use for any task about making Hex's agents trustworthy through better context—even phrased casually like “help me get my data team using AI” or “why does Threads keep picking the wrong table”. Routes to the specific workflow rather than running the full setup every time.
Generate a project-local skill that formally verifies the code where a bug costs the most and tests reach the least, in any language, with whichever verification tool fits. Use for /create-formal-verification-skill, “model-check this race”, “prove this invariant”, or when concurrency, retries, state machines, parsers, or unsafe code need more than tests.
Run GenAI evaluations on SAP AI Core: benchmark LLMs and prompts against a test dataset using system-defined and custom (LLM-as-judge) metrics, then retrieve aggregate results. Supports all three evaluation types — prompt registry, orchestration registry, and grounding (RAG) orchestration registry. WHEN: user wants to evaluate or benchmark an LLM / prompt / orchestration config, run a GenAI evaluation, create or list evaluation metrics, define a custom LLM-as-a-judge metric, create a prompt template (prompt registry) or an orchestration-registry config (with or without grounding) for evaluation, create an evaluation configuration or execution, or fetch evaluation results / scores for an execution. DO NOT USE FOR: agent evaluation via the offline aeval service — use `aeval-offline-evaluation`; building a production orchestration/RAG pipeline for inference — use `genai-hub-orchestration`; generic executions, deployments, or artifacts unrelated to evaluation — use `aicore-lifecycle-management`; credentials, resource groups, or object stores — use `aicore-admin-resources`.
Move an existing Remotion project onto Kinetograph so it renders without a browser. Use when the user has a Remotion project (remotion or @remotion/* imports, `remotion render` scripts, @remotion/three, @remotion/player) and wants it to run, preview or render with Kinetograph, or asks what in a Remotion project will not work there.
Make product launch films, promos, feature videos, logo reveals, kinetic typography and social clips with code (Remotion by default; HyperFrames, GSAP, Three.js, canvas). For a URL: capture the real logo, local fonts, colours and UI; tell a visual problem-to-payoff story, not a slideshow. Use whenever the user wants a video or animation built with code or improved, even if they only say "make a video about X" or give a URL. Optional Oneshotted library references include verified creator prompts.
Integrate the ONE store H5 Game SDK (Web SDK) into an HTML5 web game, check the integration status, and review the code before release. Use when the conversation mentions ONE store, OneStore, 원스토어, web game, 웹게임, HTML5 game, H5, WebView game, in-app purchase, 인앱결제, ads, 광고, rewarded or interstitial ads, createSDK, ONEstoreH5, initializeAsync, startGameAsync, purchase, getProductDetailsAsync, showRewardedAsync, or an iframe-hosted game, or when the user sends a command word such as doctor, integrate, iap, ads, embed, fix, or check to this skill. Covers loading the SDK, the boot flow (init, loading, start), lifecycle handling, ad and purchase code, iframe embedding requirements, error (reason) diagnosis, a blank screen in the app, and pre-release checks. When asked to write ad, purchase, or lifecycle code for a ONE store web game, use this skill instead of writing it from scratch, because it contains reference code based on the official sample.
Configure and use the SAP AI Core Orchestration service: build prompt templates, add content filtering, set up document grounding (RAG) pipelines, and call models via the Harmonized API. WHEN: user asks about the orchestration service, Harmonized API, prompt templating with variables, content filtering, grounding, RAG pipeline, document grounding, or building a pipeline that wraps an LLM. DO NOT USE FOR: creating an orchestration deployment — use `aicore-lifecycle-management` for that; credential setup — use `aicore-admin-resources`; browsing or deploying foundation models directly without a pipeline — use `genai-hub-foundation-models`; direct LLM inference without a pipeline — use `aicore-sdk-java` or `aicore-sdk-js`
Implement the confirmed requirements of the current project as actual features, screens and behavior, and perform basic checks appropriate to the implementation scope. Use for requests to write or change code, such as "build a service", "continue development", "implement the next feature" or "complete this based on these materials". Do not use when the request is only to define product direction, generate sample data, connect persistent data, fix an existing issue, validate or deploy.
Design and connect the persistent data structures and safe read/write operations required for the current project's core flows. Use when data must persist across runs, with requests such as "save the data", "connect a database", "connect Supabase" or "I want to view application records". Do not use when only static, temporary or mock data is needed.
Turn the user's business problem, idea, and supplied materials into an implementable product definition. Gradually clarify missing user, business, or UX information, whether the user has no idea, only partial context, or a clear idea with gaps. Use for requests such as “help me define what to build,” “flesh out this idea,” “create a requirements document,” or “create a PRD” when determining the user, problem, scope, acceptance criteria, and experience direction before implementation. Do not use when the product definition and UX direction are already settled and the user only requests code changes.
Reproduce and fix the cause of a current project issue based on the user's action, actual result and desired result. Use when correcting unexpected behavior, such as "I did A, but B happened", "there's an error", "it works locally but not after deployment", "the data isn't saving" or "fix it". Do not use when the user only wants to revert to a previous state without fixing the cause.
Generate safe, consistent mock sample data that matches the current project's schema and business rules. Use when the user needs to check screens, flows or tests without real data, with requests such as "make dummy data", "fill in test data" or "create demo cases". Do not use to transform real personal information or company secrets, or to produce operational data.
Set up Git, AGENTS.md, CLAUDE.md, and the initial commit point for a new project in a practically empty folder. Use for requests such as “set up the project,” “create a new project,” “prepare the workspace for a new project,” or “initialize this empty folder.” Do not use in folders that already contain code or project configuration.
Deploy a service when the user explicitly requests it, adapting to the current project's deployment environment, then verify the public URL and basic functionality. Use for requests such as “deploy it,” “make it public,” “create a URL,” “put it on Vercel,” or “let's do the first deployment.” Do not use for adding features or changing the design.
Restore code and project files to a previous working state, or reapply changes that were just reverted, while preserving current work—even if the user does not know Git terminology. Use for requests such as “revert this,” “go back to the previous state,” “undo the recent changes,” “reapply that,” or “put it back the way it was.” Do not use to fix the cause of a current problem or to restore only external data, environment variables, or deployment state.
Verify with real behavior and evidence whether the current project meets the user's request and confirmed completion criteria. Use for requests such as “Does it work properly?”, “Is it really done?”, “Test it,” “Is it ready to deploy?”, or “Recheck the changes.” Do not use for requests to change functionality or design during verification.
Works on the visuals of an edit that is already rendered, one beat at a time, without running the whole edit again. Moves, resizes, swaps, drops or adds a card, logo, real page capture or animated diagram at the moment the user names, re-plans, makes stills of just those moments, checks them (safe zone, face, motion, AI looks), patch-renders only the frames that changed and hands over on the review page. Use when the user says "add an animation here", "make a diagram of how this works", "this card is wrong", "the card covers my face", "move the card", "make the card bigger", "swap the logo", "show the real page instead", "fewer cards", "animate this screenshot", "make the motion smoother", or leaves a review note about a card. Not for the cut (cut), not for caption size, font or colour defaults (taste and style-edit), not for the first full style pass on a new cut (style-edit), and not for measuring a creator (creator-teardown).
Generate working Java boilerplate for SAP GenAI Hub using the SAP AI SDK for Java (com.sap.ai.sdk). Covers chat completion, streaming, multi-turn conversations, prompt templates, content filtering, data masking, embeddings, RAG/grounding, prompt registry, and Spring Boot streaming. WHEN: user asks for Java or Kotlin code that calls a GenAI Hub model, wants a Java starter project, asks how to use com.sap.ai.sdk in their app, or asks how to integrate AI Core into a Spring Boot or plain Java application. DO NOT USE FOR: JavaScript/TypeScript code — use `aicore-sdk-js`.
Generate working TypeScript/JavaScript boilerplate for SAP GenAI Hub using @sap-ai-sdk/orchestration, @sap-ai-sdk/foundation-models, @sap-ai-sdk/openai (native OpenAI SDK), @sap-ai-sdk/langchain (LangChain), @sap-ai-sdk/document-grounding (RAG), @sap-ai-sdk/prompt-registry, @sap-ai-sdk/llm-batch (batch jobs), @sap-ai-sdk/rpt (tabular prediction), or @sap-ai-sdk/context-registry (data destinations, tabular artifacts, scenario configurations). WHEN: user asks for TypeScript, JavaScript, Node.js, or JS/TS code that calls a GenAI Hub model, wants a TypeScript starter project, asks how to use @sap-ai-sdk in their app, asks how to use LangChain (JS/TS) with SAP AI Core, or wants JS/TS for RAG/document-grounding, prompt templates, batch LLM jobs, tabular (RPT) prediction, or managing data destinations / tabular artifacts / scenario configurations via the Context Registry. DO NOT USE FOR: Java code — use `aicore-sdk-java`.
Designs a Replay Vision scanner that produces trustworthy observations: one visible question per scanner, a type chosen from the answer shape, a query that selects only sessions able to answer it, a prompt that demands on-screen proof and allows no or inconclusive, a model picked by the cost of a wrong answer, and a first-batch test pass. Sizing and the create call stay in creating-replay-vision-scanners. TRIGGER when: user wants to design, draft, or improve a Replay Vision scanner or its prompt, asks what to scan for, asks why a scanner returns vague observations, or asks an agent to propose scanners for a product. DO NOT TRIGGER when: the question and prompt are settled and the job is to size and create (use creating-replay-vision-scanners), the scanner targets one experiment (use scanning-experiments-with-replay-vision), or the job is to read existing observations (use exploring-replay-vision-observations).
Runs diagnostics and a health check on one PostHog experiment: its setup, exposures, results and changes during the run. Covers 0 exposures, sample ratio mismatch, lost or uneven exposures, users in multiple variants, identity faults, who the experiment counts, significance traps (peeking, A/A, 'target reached', Bayesian vs Frequentist, sequential, CUPED), mid-run edits, pause and freeze, and survey follow-up. TRIGGER when: user asks 'is my experiment healthy / set up right / biased?' or 'why 0 exposures?', asks why it counts different people than their insight, mentions the bias or mismatch warning, an uneven split, significance that flips, numbers that disagree with their SQL, results that changed after an edit, or wants user feedback on an experiment. DO NOT TRIGGER when: creating an experiment (use creating-experiments), only configuring rollout (use configuring-experiment-rollout) or metrics (use configuring-experiment-analytics), or only asking lifecycle questions (use managing-experiment-lifecycle).
Explain a completed football (soccer) match from SofaScore statistics, incidents, and lineups, including beginner-friendly briefs for coaches or parents. Use for sourced post-match explanations; use multi-sport-match-research for basic cross-provider fixture checks.
Build an InboxParse email integration in a Next.js App Router app: a typed server-only client for the V1 API, a signed webhook receiver, a scheduled sync that stores every email whether or not a webhook arrived, and replies sent only after a person approves the exact text. Use when: (1) an app must receive, store, search or answer email from Gmail or IMAP mailboxes connected to InboxParse, (2) adding or auditing an InboxParse webhook endpoint, sync job or send and reply flow, (3) feeding email into an AI feature without letting the email's author steer it, (4) calling the InboxParse API directly for emails, threads, search, mailboxes, labels, contacts, webhooks or usage, (5) the user mentions: InboxParse, inboxparse.com, email API, email-to-LLM, parse emails, email webhook, X-InboxParse-Signature, whsec_, ip_ API key, INBOXPARSE_API_KEY, email.received, email.ai_processed, ai.suggested_response, /api/v1/emails, rate_limit_exceeded. Carries the facts the API reference leaves out, checked against the implementation: deliveries carry ids, never the email, are retried up to 5 times and disable the subscription after 10 failures; the signature covers the raw body with no timestamp; the list cursor pages on sent_at alone. Next.js App Router; the InboxStore interface is the seam, so Postgres, Supabase or Firestore holds the state. Not a mail server, not an email client UI, and not the InboxParse MCP server.
Links a PostHog error occurrence or session's errors to OpenTelemetry traces, logs with the same trace ID, and Session Replay. Use when the user asks for trace or span IDs on an error, same-trace logs, or backend traces for errors in a session. Starts from an occurrence UUID, session ID, or error tracking issue URL. For a general issue diagnosis or replay summary, use investigating-error-issue or investigating-replay.
Use when working with mise (mise-en-place) or planning to update mise.toml. If the project uses mise, consult it even for one-line edits to mise.toml/hk.pkl, hooks, or CI tool setup, as it has the practices and security defaults you'd otherwise skip. Trigger on the intent even when the user never says “mise” (e.g. “update/pin the node version”, “set up pre-commit”) as long as the project already uses mise. Don't load for routine mise use (run/install/trust) that isn't changing config. Optionally use the skill to transform local development into using mise, aka (mise-fy).
Make printable technical documents in the style of Apollo-era NASA paperwork. The model is an Apollo Operations Handbook "Systems Data" page with a typed running head, numbered paragraphs, Leroy-lettered line drawings and schematics, a performance data table, NOTE/CAUTION/WARNING notices, a Flight Data File checklist card, a black thumb tab and a dated footer, rendered to an A4 or US Letter PDF. Use this whenever someone asks for NASA-style, Apollo-style, retro aerospace or "mission manual" documentation. Also use it for a technical explainer, spec sheet or one-pager that should look like vintage engineering paperwork, for a handbook page explaining a product, codebase, system or fictional gadget, or for "another one of those NASA pages", even if they never say Apollo or handbook.
Build India's NHCX (National Health Claims Exchange) provider-side claims integration into an existing HMIS codebase, end to end: beneficiary policy search, coverage eligibility, the payer's insurance plan, pre-authorisation, payer queries, claim, reprocess and release, and payment notices, with the NHCX gateway (JWE, x-hcx headers, ABDM session tokens, registry, callbacks) embedded in the application. Use when adding NHCX or PMJAY cashless claims to a hospital system as a whole, or when unsure which part of NHCX is needed.
Use the Radix Community MCP server to answer questions about the Radix DLT network from the site docs, inspect any on-ledger entity (accounts, components, packages, resources, validators, transactions), and prepare transaction manifests the user signs in the web console with their Radix wallet.
Manage the full lifecycle of Alibaba Cloud Realtime Compute for Apache Flink SQL: analyze schemas, author and validate SQL, manage Drafts, Deployments, Jobs, Savepoints, Session Clusters, Catalogs, Connectors, UDFs, variables, and deployment targets, synchronize configuration to Git, and troubleshoot logs, Metrics, Checkpoints, exceptions, plans, and backpressure. Use when the user asks about Alibaba Cloud Flink, VVR, VVP, Flink SQL, its job lifecycle, or troubleshooting. Operate the platform through the Python CLI; historical Metrics use a user-supplied VVP Cookie. Verify scope and authentication before platform operations. High-risk actions usually require confirmation; explicit confirmation for the resolved target in the same request satisfies the gate.
Inspect and compare offline AI evaluation experiments, diagnose case-level regressions, and follow scorer history across application, model, or prompt changes. Use when a user asks whether an offline run improved, why a case failed, how scores changed between runs, or how to publish externally computed offline results. Covers immutable scorer versions, comparable dataset cohorts, incomplete coverage, and selective payload reads.
Provisions, inspects, updates, and deletes Google Cloud Assured Workloads V1 compliant folders, CMEK settings, Sovereign Controls by Partners, and service usage restrictions. Use when creating, listing, describing, updating, or deleting an Assured Workloads workload for any compliance regime or data boundary (FedRAMP, DoD IL2/IL4/IL5, ITAR, CJIS, Healthcare, Sovereign Controls by Partners, or Regional Data Boundaries). Don't use for brownfield assessments or custom frameworks (use assuredworkloads-environment-assessor or assuredworkloads-frameworks-*), or non-compliant folders.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.