Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
Root cause analysis on production LLM traces. Diagnoses why an LLM application is failing — works from eval judge verdicts, runtime errors, or structural anomalies depending on what signals are present. Walks the span tree from symptom to root cause. Use when user says "what's wrong with my app", "why is my eval failing", "analyze errors", "root cause analysis", "diagnose failures", or wants to understand production failure patterns.
Comprehensive web quality audit covering performance, accessibility, SEO, and best practices in a single review. Use when asked to "audit my site", "review web quality", "run lighthouse audit", "check page quality", or "optimize my website" across multiple areas at once. Orchestrates specialized skills for depth. Do NOT use for single-area audits — prefer core-web-vitals, web-accessibility, seo, or web-best-practices for focused work.
Running closed and open betas that produce real signal. Beta participant selection, structured feedback collection, beta-to-GA decision criteria, and the difference between soft-launch (no structure, no signal), kitchen-sink (everyone in, no actionable feedback), and structured beta (calibrated cohort, intentional feedback loops, clear graduation criteria). Triggers on beta program, alpha test, beta cohort, beta participant, beta feedback, beta to GA decision, design partner, early access program, closed beta, open beta, RC release. Also triggers when a feature is approaching launch and the team needs structured pre-GA validation, when prior betas produced noise rather than signal, or when the team has soft-launched before but wants more structured feedback this time.
Generates Codeception tests in PHP covering acceptance, functional, and unit testing. BDD-style with Actor pattern. Use when user mentions "Codeception", "$I->amOnPage", "$I->see", "Cest". Triggers on: "Codeception", "$I->amOnPage", "AcceptanceTester", "Codeception PHP", "Cest"
Evaluate UI designs against usability heuristics, UX laws, interaction patterns, interaction design principles, information architecture, and content quality. Conduct heuristic evaluations, structured design critiques, pattern-based reviews, posture analysis, excise audits, IA evaluations, and UX writing reviews. Provides specific, actionable feedback grounded in established principles rather than subjective opinion.
Investigate AI observability evaluations — `hog` (deterministic code-based), `llm_judge` (LLM-prompt-based), and `sentiment` (user-message sentiment). Find existing evaluations, inspect their configuration, run them against specific generations, query individual results, and set up scheduled reports on an evaluation. Use when the user asks to debug why an evaluation is failing, surface common failure modes, compare results across filters, dry-run a Hog evaluator, prototype a new LLM-judge prompt, inspect sentiment classifications, or manage the evaluation lifecycle.
Create practical ECC to S/4HANA custom-code migration dossiers for SAP packages, namespaces, object lists, or uploaded extracts. Use when asked to audit custom ABAP for S/4HANA readiness, create a shareable migration-readiness report, combine ATC + unused-code + clean-core evidence, inventory integration interfaces (RFC, IDoc, SEGW/OData, destinations) for clean core integration, classify user exits/BAdIs/enhancements/standard modifications, save results to Markdown/HTML/CSV/JSON, or visualize migration scope.
Generates TestNG tests in Java with groups, data providers, parallel execution, XML suite configuration, and listeners. Use when user mentions "TestNG", "@DataProvider", "testng.xml", "groups". Triggers on: "TestNG", "@DataProvider", "testng.xml", "TestNG suite", "parallel tests Java".
Survey an existing codebase for the structural change worth making, then rank the candidates, crown one, and write it up as a reviewable proposal. Use when the user says “where should I refactor”, “this code is hard to change”, or “audit the module boundaries”.
Comprehensive security and correctness audit of a branch's changes. Use for thermo nuclear, thermonuclear, or deep review requests, or branch/PR diff audits focused on bugs, breaking changes, security issues, devex regressions, and feature-gate leaks.
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.
Generates Flutter widget tests, integration tests, and golden tests in Dart. Supports local execution and TestMu AI cloud for real device testing. Use when user mentions "Flutter", "widget test", "WidgetTester", "testWidgets", "flutter_test", "integration_test". Triggers on: "Flutter", "widget test", "Dart test", "testWidgets", "WidgetTester", "golden test".
Bootstrap evaluators from production traces — by default propose online LLM-judge evaluators and, after you confirm, create them in Datadog as disabled drafts (never auto-enabled); on request emit Python SDK code or a framework-agnostic JSON spec instead. Use when user says "bootstrap evaluators", "generate evaluators", "create evals from traces", "eval bootstrap", "write evaluators", "build eval suite", "publish evaluators", or wants to generate BaseEvaluator/LLMJudge code or online judge configs from production LLM trace data. Works with ml_app and optional RCA report or failure hypothesis.
Analyze LLM experiment results. Handles single or comparative experiments, exploratory or Q&A modes. Use when user says "analyze experiment", "compare experiments", "analyze against baseline", or provides one or two experiment IDs for analysis.
Deterministic guardrailed agent-loop driver — wrap any step with propose → deterministic-verify → (critique → revise) and enforce hard termination guards (max_iter, wall-clock budget, no-progress via state-hash, escalate-to-human). The control loop + guards + progress detection are deterministic; the LLM critique/revise step is the one quarantined adapter. Trigger when an agent fix-loop must not spin forever and must not be able to argue its way around a failing test.
Refactoring assessment and behavior-preserving patterns for code with a passing baseline and proportionate preservation evidence. Use when the user asks to clean up, tidy, simplify, restructure, deduplicate, collapse, or merge look-alike code in a selected area, or after GREEN establishes the passing baseline for a TDD increment. Mutation testing verifies the accumulated result later at the end-of-phase PR-readiness gate. Covers recoverable-baseline discipline, when refactoring adds value vs when to skip it, and priority classification; commits always require explicit user approval. For any slice in a selected whole-path reduction program—transition or terminal—use reduce-system-complexity as the governing skill; refactoring may be secondary when applicable. For repository-wide architecture discovery use improve-codebase-architecture; for a module contract use codebase-design. Do NOT use for insufficiently evidenced code or adding behavior.
Turn wiki golden pages into a CI-blocking eval suite with a cheap→expensive assertion cascade plus a committed regression baseline. Goldens are md files under llmwiki/wiki/sources/evals/ with YAML frontmatter (input, expected, optional rubric, asserts). The engine runs deterministic tier-1 asserts (equals/contains/regex/is-json/is-sql-ish) in pure code, an optional tier-2 similarity, and reports tier-3 (LLM-rubric judge) as the quarantined adapter — never calling a model. Candidate outputs come from --outputs outputs.json so the suite is fully runnable without any model. Trigger when the user says "wikieval", "eval suite for the wiki", "regression gate for goldens", "assertion cascade", "block CI on eval drop", or invokes /wikieval.
Expert iOS App Store submission and approval system. 9 specialized agents provide senior App Review Team-level expertise across compliance, design, privacy, monetization, metadata, technical requirements, timing, rejection recovery, and learning. Triggers on keywords like app store, iOS submission, apple review, app rejection, aso, privacy manifest, privacy labels, ATT, iap, in-app purchase, subscription, storekit, review guidelines, HIG, testflight, app store connect.
Expert in browser automation using Chrome DevTools Protocol (CDP) and WebDriver. Specializes in secure web automation, testing, and scraping with proper credential handling, domain restrictions, and audit logging. High-risk skill due to web access and data handling.
Root-causes AI agent failures from MLflow traces without requiring a code change, regression suite, scorer, or evaluation run. Use when the user asks to debug, investigate, explain, or determine why an agent produced an output. Prefer an existing matching trace; reproduce only when no usable trace exists.
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.
This skill should be used when the user asks to “test for XSS vulnerabilities”, “perform cross-site scripting attacks”, “identify HTML injection flaws”, “exploit client-side injection vulnerabilities”, “steal cookies via XSS”, or “bypass content security policies”. It provides comprehensive techniques for detecting, exploiting, and understanding XSS and HTML injection attack vectors in web applications
End-to-end LLM Observability pipeline for an instrumented ml_app — classify production traces, root-cause failures, bootstrap evaluators, then (optionally) sample + publish a dataset, generate + run an experiment, and analyze results. Six narrated phases with a standardized banner and a “continue” checkpoint between each. Pure orchestration over the dd-llmo sub-skills (`llm-obs-session-classify`, `llm-obs-trace-rca`, `llm-obs-eval-bootstrap`, `llm-obs-experiment-py-bootstrap`, `llm-obs-experiment-analyzer`). Use when user says “run the eval pipeline”, “go from traces to evals”, “bootstrap evals end to end”, “classify then RCA then bootstrap”, “build an eval set from scratch”, “onboard me to datasets and experiments”, “walk me through experiments”, “I have an ml_app, now what”, “LLM Obs onboarding”, “guided experiment setup”, “from traces to experiments”, or wants a deterministic, narrated tour from production data through evaluators, datasets, and experiments. Stop early with `--stop-after <phase>` to short-circuit at evaluators or dataset, or resume mid-flow with `--start-at <phase>`
Reviews whether a design is SOUND — solving the right problem, derived from its stated goals and constraints — and emits ranked, evidence-backed findings, not edits. Read when asked to “review this design”, “is this design sound”, “pressure-test this proposal”, “do a design review”, “does this solve the right problem”, “poke holes in this spec”, “should we build this”, or to critique a proposal / spec / ADR / architecture or product decision. Do NOT read when the user wants to AUTHOR one of these — routing a new proposal is frame-a-proposal, a spec is write-a-spec, a decision record is record-a-decision, a postmortem is write-a-postmortem. Do NOT read for code review of a diff, or to fact-check individual claims (that is a correctness pass, a different job)
Generates SmartUI visual regression test configurations for screenshot comparison on TestMu AI cloud. Framework-agnostic—works with Playwright, Selenium, Cypress, and Puppeteer. Use when the user mentions “SmartUI,” “visual regression,” “screenshot comparison,” or “visual testing.” Triggers on: “SmartUI,” “visual regression,” “screenshot comparison,” “pixel diff,” “visual testing LambdaTest.”
Configure a GitLab CI/CD pipeline that embeds SAST (Semgrep, SpotBugs, Gosec, Bandit, NodeJsScan), DAST, container scanning, dependency scanning, and secret detection via GitLab's managed security templates. Use when building a shift-left DevSecOps pipeline in GitLab, adding automated vulnerability scanning stages to .gitlab-ci.yml, or triaging scanner findings with GitLab Duo AI before deployment.
Generate connectivity knowledge for a SaaS API when no dedicated Mule connector exists. Researches the API based on user-defined use cases and documentation, produces an OpenAPI 3.0 specification, asks the user for credentials, validates every operation against the live API with auto-fix, and writes connectivity-schema/<apiName>/ (api-reference.md + <apiName>.yaml + config.properties). The resulting folder feeds downstream HTTP-Connector flow generation in build-mule-integration so an HTTP fallback path inherits the same auth, pagination, and entity awareness a dedicated connector would carry. Use when the user's prompt names a SaaS for which no Exchange connector was found, or asks to “generate API knowledge,” “research a SaaS API,” “build connectivity schema,” “validate spec against live API,” or similar
Godot expert auditor: Aurelius. Exhaustive never-list enforcement and architectural slap-down for Godot 4.7 projects. Use when auditing signal decay, ObjectDB orphans, typed Array/Dictionary slop, material.duplicate batch breaks, export case-sensitivity, Expression.execute risks, or sector never-lists. Keywords: auditor, Aurelius, never-list, signal decay, ObjectDB orphans, typed Array, export case-sensitivity, instance uniforms, PackedScene.get_state.
Investigate a GitHub issue — fan out parallel exploration, find the root cause (5 Whys, evidence-backed), and write a reviewable RCA artifact (then post a summary to the issue). The investigate step before piv-implement-issue. Use to diagnose a bug/issue before fixing it.
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.