Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
Guide users through targeted manual verification after code changes. Use when asked to "test this", "verify it works", "QA this", "walk me through testing", "smoke test", "sanity check", "regression test", "acceptance test", or after implementing a feature or bug fix that still needs human validation. Favor this skill for focused verification of the current change; use a broader exploratory-testing skill for open-ended bug hunting across an entire app.
(NS) Diagnose a product bug report (screen, expected vs actual, ticket, screenshot description) by tracing the reported flow across every layer the consumer stack has. Chat-only eight-section report, then auto-dispatch ns-execution-gitlab-issue (ISSUE_URL) or ns-coder unless the user opted out (diagnose-only / do not fix). This skill does not write the patch. Use whenever the user pastes a relato, issue, ticket, screenshot description, "what is happening", "diagnose this", expected vs actual on a screen, or /ns-bug-report-diagnosis — even if they never name this skill. Do NOT use if they asked to implement/fix now without a screen story (ns-coder), to execute a GitLab ISSUE_URL as the first skill (ns-execution-gitlab-issue), for stack/CI/log with no screen story (ns-investigator), or for DB-only inspection when the consumer has a DB skill
Use proactively when a task requires implementing, modifying, refactoring, or reviewing a product's frontend visible interface, interaction behavior, frontend state and data flow, component or style organization — no need for the user to name it. Also for testing and validation strategies for these frontend behaviors. Applies to pages, forms, lists, modals, responsiveness, styling and motion implementation, and frontend code ownership adjustments. Not for design judgments such as visual style, color scheme, typography, and design direction, nor for purely backend work, build and deployment, dependency upgrades, mechanical changes that only fix import paths, merely operating an existing website, only generating design images, or discussing this skill; not triggered when the user explicitly asks not to use it.
Use when optimizing, auditing, or preparing to ship web animations or rendering performance work: frame loops, scroll/viewport reveals, mount/unmount transitions, canvas/WebGL lifecycles, reduced motion, lazy rendering, and deferred off-screen work. Use for jank, per-frame allocations, forced reflows, render loops, animations that do not pause off-screen, content-visibility, requestIdleCallback, analyzing Chrome DevTools performance traces for animation or rendering issues, or choosing between browser-driven CSS/WAAPI, minimal JS, phase, and Motion/GSAP. Use when the user explicitly asks about phase APIs or behavior. Do not use during exploratory animation ideation, prototyping, visual iteration, or trying variants, even when the code already imports phase; wait until the user asks for performance guidance, an audit, production implementation, ship-readiness review, or about phase itself.
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
Add or verify CI enforcement of Conventional Commits pull request titles. Detects whether the repo squash-merges with the PR title as the commit subject, checks for an existing title check, scaffolds `.github/workflows/pr-title.yml` using amannn/action-semantic-pull-request at a pinned SHA, and verifies it fails a bad title and passes a good one. Use when the user says enforce conventional commit PR titles, lint PR titles, semantic pull request check, add a PR title check, or invokes /pr-title-lint. Do NOT use for switching a repo to squash-merge with PR-title commit messages (use gh-bootstrap for merge defaults and rulesets), cutting a release (use release), or local commit-message hooks (use prek).
When the user wants to improve product quality, implement quality control systems, reduce defects, or analyze quality data. Also use when the user mentions "SPC," "statistical process control," "Six Sigma," "DMAIC," "control charts," "process capability," "Cp," "Cpk," "quality control," "defect reduction," "inspection," "acceptance sampling," or "quality assurance." For lean improvements, see lean-manufacturing. For process optimization, see process-optimization.
Run a read-only audit of every catalog skill's checks against a repository and report which setup gaps exist. Runs a bundled Python script from the repository root, prints one row per catalog skill (check, status, evidence, the skill that fixes it), then proposes an ordered plan naming those skills. Never writes a file and never calls a mutating GitHub API endpoint, even with the optional --github flag, which adds read-only GET checks. Use when the user says "audit this repo", "what's missing in this repository", "repo health check", "which setup skills apply", or invokes /repo-audit. Do NOT use for applying any fix directly — route each finding to its catalog skill — or for code review of application logic.
Audit whether a result can actually be reproduced from the available code, data, configs, and documentation. Use before trusting a benchmark claim or releasing your own work. Use for replication, code release checks, paper artifact sanity checks, lab handoffs, catching missing seeds/configs/data before results are cited, or whenever "works on my machine" is a risk. If the user wants to plan what to include in a public artifact package, use artifact-release-packager instead; this skill's job is to test whether reproduction is possible and where it breaks.
Inspect, diagnose, fix, and verify responsive frontend defects across routes, states, containers, and input modes. Use for mobile or desktop QA, overflow, overlap, broken navigation, unstable grids, clipped text, blank media, viewport regressions, screenshot comparison, and post-implementation visual verification.
Use when adding, changing, testing, or debugging Rust HTTP APIs and services, especially when Codex needs black-box integration tests, random-port app startup, real database test isolation, external HTTP mocks, or CI-ready cargo verification for Actix, Axum, Warp, Rocket, or similar Rust web frameworks.
Use when a Roblox experience is slow, laggy, stuttering, crashing or out of memory, or before shipping or updating one: my game lags on mobile, low FPS, frame spikes, server lag or low heartbeat, physics lag, network lag or high KB/s, memory leaks or memory that keeps growing, crashes on phones, long join or load times, too many parts, instances, draw calls or triangles, MicroProfiler, Script Profiler, Luau heap, Developer Console memory, Scene Analysis, StreamingEnabled tuning, performance budgets, and publishing, releasing an update, restarting servers or rolling back a version.
Use when setting up or repairing a Roblox team project outside Studio: Rojo or Script Sync, Wally packages, Rokit pins, luau-lsp type checking, Selene, StyLua, unit tests in Lune or the engine, GitHub Actions CI on pull requests, Open Cloud place publishing, Luau Execution or API key scopes, merging builders' Team Create map with code from files, Studio plugins, or proving a pipeline works. Also when users say "set up Rojo", "CI for my Roblox game", "run tests in CI", "publish from GitHub", "Script Sync stopped syncing", "Rojo deleted my parts", "wally packages not found", "luau-lsp can't find ReplicatedStorage", "selene errors on types", "the release wiped the map" or "my LocalScript doesn't run".
Act as a Senior Backend Developer expert to design system architectures, build production-grade APIs (REST, GraphQL, gRPC), write optimized database schemas and queries, perform rigorous code reviews, and enforce engineering best practices. Supports Node.js/TypeScript, Python, Java/Spring Boot, Go, and Rust. Produces production-ready code, architecture docs with Mermaid diagrams, migration files, and actionable code-review feedback. Use whenever a user needs backend engineering guidance, code generation, refactoring, or architectural decisions.
Design behavior evals, positive and negative prompts, expected behaviors, rubrics, fixtures, golden outputs, forward-tests, regression cases, and quality gates for AI agent skills and reusable agent workflows. Use when creating or improving SKILL.md packages, validating trigger descriptions, proving agent behavior changes, avoiding prompt dumps, or building eval-backed skill marketplaces.
SQLAlchemy ORM and Alembic migration best practices for building safe, performant database schemas. This skill should be used when writing, reviewing, or refactoring SQLAlchemy models, Alembic migrations, or database query patterns. Triggers on tasks involving SQLAlchemy ORM, Alembic migrations, database schema changes, or query optimization.
Review test quality and audit test coverage for any module. This skill should be used when reviewing existing tests, auditing test gaps, writing new tests, or when asked to assess test health. It pipelines testing standards into the audit workflow to produce a prioritized gap report. The output is a report, not code — do not write test implementations until the report is reviewed.
Use this skill to test `LazyColumn`, `LazyRow`, and `LazyVerticalGrid` correctly — tag the container with `Modifier.testTag(...)`, tag each item by its key, scroll via the semantic action (`hasScrollAction()`, `performScrollToIndex`, `performScrollToKey`) or via `Modifier.testTag` plus `performTouchInput { swipeUp() }`, and verify visibility with `assertIsDisplayed` (NOT `assertExists`, since off-screen lazy items still exist as semantic nodes in some configurations). Covers `LazyListState.layoutInfo.visibleItemsInfo` as the highest-signal probe, `mainClock.autoAdvance = false` for animated item placement, and the parameterized vertical/horizontal base-class pattern. Use when the developer asks "how do I scroll a LazyColumn in a test", "assertExists vs assertIsDisplayed", "scroll to a key", "test item placement animation", or sees a test pass when an item is off-screen.
Use when creating or editing a skill that enforces discipline, before deploying it, and when an existing skill is being ignored, argued with, or followed only when convenient. Triggers include writing a new skill, changing a skill's rules, discovering an agent found a loophole in guidance, and needing to know whether a skill actually changes behavior rather than just reading well.
Use this skill to size an Android test suite using Google's small / medium / big scope vocabulary and the qualitative pyramid. Explains why most apps should hold "many small tests and relatively few big tests", how scope (small/medium/big) is orthogonal to execution location (local vs instrumented), and where the older 70/20/10 numeric ratio actually comes from. Use when the user asks "how many unit vs UI tests should I write", "what's the testing pyramid", "all my tests are instrumented and CI is slow", "test pyramid 70 20 10", "Android small medium large tests", "Robolectric counts as which layer", or mentions Unit / Component / Feature / Application / Release Candidate framing from the strategies page.
Validate in any of six modes — spec quality, single-file review, spec-to-implementation drift, constitution rule enforcement, comparison between two artifacts, or sanity-checking your understanding. Use when checking a spec by ID, validating a file by path, detecting drift, enforcing constitution rules, comparing two sources ("$X against $Y"), or asking a freeform validation question.
Use this skill when you need to verify cross-file data/asset consistency in a content-driven site — not "does it render?" (that's `web-visual-verification`) but "do the data, files, and references all line up?". Triggers on phrases like "盤點內容", "稽核資產", "對照 course-data 跟 markdown", "找缺圖", "task ID 有沒有重複", "quiz 編號 vs 硬編碼總數", "audit", "content audit", "content drift", "asset coverage", "three-way sync check", "cross-file consistency", "find missing illustrations", "資料一致性檢查", "找出該補插圖的地方", or any moment when the user senses divergence between source files and deployed data. Output is a human-readable markdown report, not a pass/fail. Pair with `web-visual-verification` for full pre-release coverage.
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.