Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
(NS) Root-cause diagnosis from errors, logs, stack traces, failing tests, or CI — entry priority 4b when the user wants diagnosis WITHOUT implement ("why does X break?"). Use proactively on unclear test/CI failures if they have not asked to fix yet. Do NOT use for implement/fix (ns-coder), product bug reports / screen expected-vs-actual (ns-bug-report-diagnosis), version scope (ns-spec-driven), GitLab ISSUE_URL execution, or pre-merge review (ns-reviewer).
Review a Penpot design against its implemented code (a source component or Storybook story) and emit a structured DRIFT report. Use when an engineer needs to confirm the build matches the design: extract HTML/CSS from a selection, map its tokens/structure/states to the code component, diff them, and list what each side is missing with severity and a reconciliation. Degrades gracefully to a design-system-token check when no code source is supplied. Triggers: 'design to code review', 'does my code match the design', 'design code parity', 'compare Penpot to my component', 'check this against Storybook', 'extract HTML/CSS from this selection', 'find design drift', 'design implementation diff'.
Practical Python craftsmanship guidance based on One Python Craftsman. Use when writing, refactoring, or reviewing Python code for naming, branching, data structures, functions, exceptions, loops, decorators, imports, file I/O, edge cases, and modern syntax choices. If the skills set includes friendly-python, suggest invoking it for better Python outcomes.
Generates production-grade Reqnroll BDD automation scripts for web (Selenium 3/4) and mobile (Appium 2) testing in C#. Supports parallel NUnit execution locally and on TestMu AI cloud. Use when the user asks to write BDD tests, automate with Reqnroll, create .feature files, write Gherkin scenarios, write step definitions, migrate from SpecFlow, or test on browsers/Android/iOS. Triggers on: "Reqnroll", "BDD", "Gherkin", ".feature file", "step definition", "SpecFlow migration", "Selenium C#", "Appium C#", "TestMu", "LambdaTest", "NUnit BDD", "reqnroll.actions.json".
Fetch, analyze, reproduce, and minimize GitHub issue reproductions. Use only when the user asks to check whether a GitHub issue reproduces, minimize a GitHub issue's repro, analyze a GitHub bug report, or scrub/triage a GitHub issue for reproducibility. Do not use for standalone reproduction requests without a GitHub issue.
Drive and inspect the Umbraco backoffice reliably with Claude-in-Chrome browser automation. Use this WHENEVER you are about to use the mcp__claude-in-chrome tools against a running Umbraco backoffice — navigating sections, clicking tree items, filling forms, opening/closing modals, or visually verifying a backoffice extension (property editor, dashboard, workspace, tree, section, menu). The backoffice is Lit/web-components with open shadow DOM and modals that render off-screen, so the built-in find/read_page tools see nothing and screenshots lag behind the live DOM. This skill provides a shadow-DOM-piercing JS helper toolkit plus a navigation playbook so you read state from the DOM instead of guessing from pixels. Trigger even when the user only says "open the backoffice", "test the extension in Umbraco", "click through the backoffice", "log in and check it works", or "take a screenshot of the editor in Umbraco".
Generate Bruno API collections, request files, environments, and request documentation from backend source code, route definitions, OpenAPI snippets, pasted API docs, curl commands, or endpoint lists. Use when a user asks to create, scaffold, convert, organize, or improve Bruno collections, opencollection YAML files, .bru files, folders, request docs, request metadata, environments, or API-client collections for Bruno.
Automated code review for Arduino/ESP32/RP2040 projects focusing on best practices, memory safety, and common pitfalls. Use when user wants code feedback, says "review my code", needs help improving code quality, or before finalizing a project. Generates actionable checklists and specific improvement suggestions.
DDD code review and anti-pattern detection — anemia model detection, layered compliance checking, aggregate design review, rich model verification, scoring system (5 dimensions), and ArchUnit automated validation. Use when user asks about code review, 代码审查, DDD review, 架构审查, anti-pattern detection, 反模式检测, or needs to check if code follows DDD.
Comprehensive GTM tracking testing and validation including automated Playwright headless testing, browser console testing, GTM Preview mode validation, and GA4 DebugView verification. Use when users need to "test GTM tracking", "validate dataLayer events", "debug GTM", "check if tracking works", "automated tracking tests", "run tracking tests without opening browser", or troubleshoot tracking issues. Prioritises automated testing over manual when possible.
Transition from vibe coding (fast prototyping) to agentic engineering (rigorous iteration). Use this skill when the user has a rough prototype and wants to level it up, needs to go from "it kind of works" to production-quality, wants to understand the difference between vibe coding and agentic coding, or says "polish this", "make this production-ready", "level up from prototype", "vibe coded this now what". Based on Karpathy posts 3 and 6.
Run the konsistent CLI, review all reported violations, and fix them in the codebase. Use when the user wants to "fix konsistent violations", "resolve konsistent errors", "clean up konsistent diagnostics", "make konsistent pass", "address structural convention violations", or asks to run konsistent and act on the results. The skill is interactive: it requires explicit user decisions when many violations stem from the same rule (rule vs. code question), explicit user confirmation before modifying any code, and a triage step for non-trivial violations. For editing `konsistent.json`, this skill defers to the `konsistent-config` skill.
Internationalization (i18n) and localization (l10n) testing for global products including translations, locale formats, RTL languages, and cultural appropriateness. Use when launching in new markets or building multi-language products.
Automated PCI Secure Software Standard (SSS) v2.0 readiness gap-assessment of an application from its source code and documentation. Deterministically enumerates every applicable Test Requirement from a pinned catalog, gathers source/doc evidence, and emits an evidence-bound per-requirement verdict (MET / NOT_MET / PARTIALLY_MET / NOT_APPLICABLE / REQUIRES_MANUAL_REVIEW) with file+line citations, then a Transilience gap report. Use when asked to assess, gap-assess, or check an application's readiness against the PCI Secure Software Standard v2.0 / PCI SSF, or to map source code and design docs to PCI SSS Security Objectives and Test Requirements. Produces a readiness gap-analysis only — NOT an official PCI validation.
Audit a Penpot design against WCAG 2.1/2.2 AA (optionally AAA) and produce a structured severity report: color contrast, touch/target sizes, text alternatives, heading hierarchy, and keyboard/focus order. Proposes fixes — does not auto-apply. Also serves as the Evaluator in the brief-to-screen and brief-to-deck loops (scope:deck = slide contrast rules). Triggers: 'check accessibility', 'accessibility audit', 'WCAG check', 'contrast audit', 'a11y review', 'check color contrast', 'is this accessible'.
Use when writing E2E tests with Playwright, setting up test infrastructure, debugging flaky browser tests, organizing project structure, or testing REST APIs. Invoke for browser automation, E2E tests, Page Object Model, test flakiness, visual testing, project scaffolding, folder layout, API testing, JSON schema validation.
Review requirements before development. Identify ambiguity, missing information, contradictions, unclear acceptance criteria, logical gaps, edge cases, and testability issues. Determine whether the requirements are sufficiently clear for development and QA. Use this skill when a user provides requirements (BRD, user story, use case, feature ticket, specification, or prose) and asks for feedback, validation, review, or readiness assessment before implementation.
Assess and respond to pull request review comments through a status-gated workflow: collect comments, evaluate feedback, draft replies, write a verified local report, and optionally post exact approved replies to supported GitHub review-comment threads.
Review Rust changes for correctness, memory and thread safety, error semantics, unnecessary allocation or cloning, lock scope, API compatibility, test gaps, documentation, and dependency risk, applying the Rust API Guidelines checklist (C-PANIC, C-UNWRAP, C-TRANSMUTE, C-BOOL, C-NEWTYPE, C-COMMON-TRAITS, C-CONVERT, C-SEALED, C-NON-EXHAUSTIVE). Use when reviewing Rust diffs, pull requests, libraries, unsafe boundaries, or production incidents; report actionable findings by severity before summaries, route automated formatting or lint policy to rust-style-clippy, and route API shape decisions (trait sealing, error taxonomy, newtype design, builder patterns) to rust-api-design.
Activate for the daily repository triage run, or when the user asks to triage CI failures, open issues, or recent commits. Reads yesterday's signals, dedupes against the state file, drafts fixes in isolated worktrees with a reviewer pass, opens PRs on green, and keeps people out of limbo with comments in the maintainer's voice. The state file is the spine: it remembers what was tried, what passed, and what is still open, so each run resumes where the last one stopped.
Orchestrated, holistic UX review of a screen, flow, or component. Establishes context once, auto-runs the always-relevant lenses (usability, microcopy, dark patterns), offers the context-dependent ones (accessibility, aesthetics, IA, mobile, visual foundations), and returns one prioritized report. Trigger when the user asks for a full design review, "review my UI/screen/flow", "is this good UX", "audit this design", or general design feedback — but NOT when they name a single specific lens.
Validate that an Antithesis property catalog can actually catch bugs. For each safety property, inject one realistic mutant into the SUT, run it, and confirm the targeted property fires. Diagnoses survivors as a bad mutant, bad oracle, bad workload, or bad property, and routes fixes back to antithesis-workload or antithesis-research. Use once the harness is built and a baseline run is green.
This skill should be used when the user asks to "audit this codebase", "audit this code", "code audit", "check for bugs", "review code quality", "find dead code", "check for anti-patterns", "performance audit", "check for code smells", "technical debt", "code health check", "review test quality", or "check error handling". For security-focused audits (vulnerabilities, OWASP issues, injection, authentication or authorization weaknesses, secrets handling, cryptography review, dependency CVEs, IaC, CI/CD), use the `security-audit` skill instead.
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
Review code for bugs, logic errors, security vulnerabilities, code quality issues, and adherence to project conventions, using confidence-based filtering to report only high-priority issues that truly matter. Use this skill when reviewing a small set of changes locally (such as unstaged diff), when dispatched as a sub-task during feature-dev quality review, or when the user wants a critique of a specific file or function.
Reviews Deep Agents code for bugs, anti-patterns, and improvements. Use when reviewing code that uses create_deep_agent, backends, subagents, middleware, or human-in-the-loop patterns. Catches common configuration and usage mistakes.
Remove AI slop from the code a branch changed, before a commit and without changing behavior: narrating comments, guards on states the types or the boundary already rule out, dead compatibility paths, unrelated edits, one-caller wrappers and speculative abstractions, casts that silence the type checker, and nesting an early return flattens. Use for /deslop, deslop it, clean the diff, strip the slop, or before any commit. Code only. The unslop skill cleans prose and the no-comments skill reviews comments.
Use this skill when the user measures control-plane insertion throughput of DOCA Flow with the doca_flow_perf binary on x86 host or Arm cores—selecting a configs/ JSON policy, weighing the DPDK-versus-DOCA backend, running the smoke iteration then the eval loop, interpreting CPU cycles plus num_pushed / num_failed, or capturing the four-tuple (DOCA release, firmware, JSON policy, worker config) that makes an insertion figure defensible. Also load it for implied asks like "5-tuple hairpin insert figure", "flow-perf result mismatches release notes", "DPDK-vs-DOCA comparison", or "install variance too high". Refuse and route to the proper owner for live-app optimization (doca-flow-tune), wire throughput or latency, or library-internal pipe semantics.
Scan generated/authored HTML for accessibility violations with axe-core (Deque) running over JSDOM — structural and semantic rules: ARIA attributes and roles, accessible names, alt text, form labels, heading order, landmarks, semantic HTML, tabindex, duplicate IDs, lang attribute, and ~50 more. Use on the CODE side of a Figma-to-code workflow: triggers 'scan this HTML for accessibility', 'run axe on my component', 'check the generated code for a11y', 'axe-core scan', 'accessibility lint my markup', 'WCAG check on code', 'is this HTML accessible'. This is a standalone Node script — NO Figma connection required. Visual rules (color-contrast, focus-visible, link-in-text-block, target-size) are DISABLED because JSDOM has no layout — use figma-lint-design for visual a11y on the design side. The two together give full design+code coverage.
Build practice environments and evaluation gyms where agents can try, fail, and learn from feedback. Use this skill when the user wants to set up automated evaluation for an agent, needs a sandbox for testing LLM capabilities, wants to build a practice gym for skills, needs an RL-style environment for agent improvement, or says "eval environment", "practice gym", "agent sandbox", "automated testing for agents", "RL environment", "agent eval loop". Based on Karpathy LLM Textbook and environments posts.
Take your AI agent to the next level with full LangWatch integration. Adds tracing, prompt versioning, evaluation experiments, and simulation tests in one go. Use when the user wants comprehensive observability, testing, and prompt management for their agent.
Blind side-by-side multi-model comparison. Send one prompt to 2-4 models simultaneously, present responses anonymously (Model A / B / C / D), let the user pick a winner, then reveal identities and show which model won. Supports custom evaluation criteria, synthesis of responses, and vote history logging. Trigger when the user says "compare models", "test these models", "which model is better for", "A/B test", "blind comparison", "model evaluation", or wants to see how different AI models handle the same prompt. Can also be used for prompt engineering — testing how different models interpret the same instructions. Supports reasoning-effort A/B via provider:@effort:model_id specs.
(NS) PHPUnit unit/integration in backend package (bootstrap tests/ when missing). Docker-only: phpunit --testdox --stop-on-failure --stop-on-error, 120s. Use when writing/refactoring PHPUnit, unit-test tasks, or scaffolding backend tests on greenfield Laravel/PHP — not when planning test task markdown (ns-spec-driven references/unit-test-task-generator.md).
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.