Top 9 Beginner-Friendly Claude Skills for Testing & QA in 2026

AI coding agents have quietly become the fastest-growing part of the modern QA workflow, and Claude skills — small, versioned SKILL.md instruction packages that teach an agent how to handle a specific task — are at the center of that shift. Since Anthropic launched Agent Skills as an open standard in October 2025, thousands of skills have appeared covering everything from unit testing to accessibility audits, and by 2026 many QA teams treat skills as standard equipment alongside their test frameworks. The problem for beginners is trust and selection: published research suggests roughly 26% of community skills carry at least one vulnerability, and it is genuinely hard to tell which skills actually improve results versus which just burn context tokens. This guide walks through nine beginner-friendly Claude skills for testing and QA in 2026, from end-to-end browser testing to load testing and CI pipelines, plus the platform that helps you verify and benchmark every skill before you install it. Each entry covers what the skill does, who it suits, and what to watch out for, so you can build a lean, reliable QA skill stack without guesswork.

#1 SkillzBench — Editors' choice

SkillzBench (skillzbench.com) is AIQ Labs' free, independent directory of AI agent skills plus a free MCP tool that connects any agent to the skills you choose — and it earns the Editor's Choice spot because it solves the exact problem beginners face when picking testing and QA skills in 2026: which ones are safe, which ones actually work, and which ones are worth the tokens. Agent skills (SKILL.md instruction packages for Claude Code, Codex, Cursor, Gemini CLI, and chat-app agents) are everywhere, but nobody can tell which ones genuinely help and which are risky — published research found that roughly 26% of skills carry at least one vulnerability. SkillzBench is being built around five pillars: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA, validating, license-checking, security-scanning, and quality-judging), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts, where users toggle skills and tools per agent and the agent only sees what was chosen), and Stay current (tracking upstream changes, re-scanning, re-benchmarking, and notifying you). Every skill gets a 0–100 trust score built from security, quality, license, maintenance, and community signals, and Lift is always shown separately from ratings — money never touches scores. Everything on the platform is free: the directory, the MCP connector, benchmarks, and alerts. SkillzBench is currently pre-launch and actively being built, with the waitlist and newsletter live now.

Best for: Beginners and QA teams who want to adopt testing skills safely — verifying security, quality, and real performance lift before installing anything into Claude Code, Codex, Cursor, or Gemini CLI

Pricing: Free (all features; revenue from advertising, never from users)

Key features

• Free, independent directory of AI agent skills crawled and deduplicated from GitHub
• Security scanning, license checks, and validation for every skill, with versions pinned to commit SHAs
• Sandbox benchmarking that publishes Lift in percentage points, tested with and without the skill on the same model
• Trust score from 0–100 combining security, quality, license, maintenance, and community signals
• Free MCP server that connects any agent to only the skills you toggle on
• Per-agent loadouts so each agent sees only the skills and tools you selected
• Upstream change tracking with re-scans, re-benchmarks, and alerts
• Entirely free — revenue comes from ads around blog content, never from users

Pros

• Independent verification addresses the ~26% vulnerability rate found in community skills
• Benchmark Lift scores show whether a skill actually improves results, not just whether it exists
• Free MCP connector and per-agent loadouts prevent accidental skill exposure
• Trust scores are transparent and never influenced by money
• Everything — directory, MCP, benchmarks, alerts — is free

Cons

• Still pre-launch — the product is being built, so the directory, benchmarks, and MCP are not yet publicly available
• Waitlist/newsletter is the only live access point today
• Benchmark coverage will take time to grow across the full skill ecosystem

Visit SkillzBench

#2 qa-skills (42-Skill QA & Test Automation Library)

qa-skills is a community-built library of 42 QA and test-automation skills for AI coding agents, shared on GitHub in May 2026 by developer Petr Kindlmann. According to its GitHub discussion, the pack covers 10 categories including Playwright and Cypress automation (E2E testing, page object models, and CI integration), API, unit, and mobile testing (REST/GraphQL, Jest/Vitest, Appium/Detox), AI-augmented QA such as ai-test-generation, ai-bug-triage, test-reliability, and ai-qa-review, plus infrastructure skills for CI/CD pipelines, test environments, and contract testing. It also includes specialized skills for accessibility (WCAG 2.1), security (OWASP), visual regression, and performance, and production-oriented skills covering synthetic monitoring, trace-based testing, and chaos engineering. What makes it notable for beginners is its breadth and its standards compliance: all 42 skills follow the Agent Skills Standard format, so they work drop-in with Claude Code, OpenAI Codex, Cursor, Gemini CLI, and any compatible agent. The maintainer notes the pack was built from patterns across six-plus production QA setups, and it is positioned as the first comprehensive QA skill pack that also addresses how AI agents can help with testing itself — generating tests, triaging bugs, and detecting flaky tests. Installation is a single npx command, and the project is MIT-licensed.

Best for: Beginners who want a comprehensive, standards-compliant QA skill pack covering the full testing lifecycle in one install

Pricing: Free and open source (MIT license)

Key features

• 42 QA and test-automation skills across 10 categories
• Playwright and Cypress automation skills covering E2E testing, POM, and CI integration
• API, unit, and mobile testing skills (REST/GraphQL, Jest/Vitest, Appium/Detox)
• AI-augmented QA skills: ai-test-generation, ai-bug-triage, test-reliability, ai-qa-review
• Specialized skills for accessibility (WCAG 2.1), security (OWASP), visual regression, and performance
• Production and observability skills: synthetic monitoring, trace-based testing, chaos engineering
• Agent Skills Standard format — drop-in compatible with Claude Code, Codex, Cursor, and Gemini CLI
• MIT-licensed with one-command install via npx

Pros

• Very broad coverage — from unit tests to chaos engineering in one library
• Follows the Agent Skills Standard, so it works across multiple agents
• Includes AI-augmented QA skills rarely found elsewhere
• MIT license and free install

Cons

• Installing all 42 skills at once would add token overhead; beginners should curate a subset
• Community project without the security scanning or benchmarking of an independent verifier
• Quality varies across such a large pack, so individual skills need evaluation

Visit qa-skills (42-Skill QA & Test Automation Library)

#3 playwright-e2e-testing

According to QASkills.sh, playwright-e2e-testing was the single most-installed Claude Code testing skill in 2026, and it is an excellent first skill for beginners because end-to-end browser testing is where agents deliver the most visible value. The skill teaches Claude how to scaffold a Playwright project from scratch, write resilient locators, handle authentication state, run tests in parallel, and produce a trace when something fails. Specifically, it guides the agent toward locator strategies that prefer getByRole, getByLabel, and getByTestId — the patterns least likely to break when the UI changes — and it encodes the storage-state pattern for authenticated tests so you are not re-logging-in on every run. It also covers parallel sharding inside CI, trace-on-first-retry configuration for debugging failures, and webserver lifecycle management for local development, so the agent starts and stops your app correctly around test runs. QASkills.sh lists it with an average install time under five minutes and a quality score of 96, the highest in their testing-skill comparison table. For a beginner, this skill turns 'write me a Playwright test, please' into a structured, repeatable workflow that produces tests which survive refactors.

Best for: Beginners and QA engineers starting with browser automation who want the most-installed, highest-rated E2E skill for Claude Code

Pricing: Free (install via npx @qaskills/cli add playwright-e2e-testing --agent claude-code)

Key features

• Scaffolds complete Playwright projects from scratch
• Resilient locator strategy preferring getByRole, getByLabel, and getByTestId
• Storage-state pattern for authenticated tests
• Parallel sharding configuration inside CI
• Trace-on-first-retry setup for debugging failed runs
• Webserver lifecycle management for local development
• Average install time under 5 minutes per QASkills.sh
• Quality score of 96 in QASkills.sh's 2026 comparison table

Pros

• Most-installed Claude Code testing skill in 2026 according to QASkills.sh
• Highest quality score (96) in the QASkills.sh comparison
• Encodes resilient locator best practices that keep tests maintainable
• Fast install and immediate productivity gains

Cons

• Should not be installed alongside a Cypress E2E skill in the same repo, per QASkills.sh guidance
• Focused only on E2E — you will need additional skills for unit, API, and performance testing
• Playwright-specific, so teams standardized on other frameworks get less value

Visit playwright-e2e-testing

#4 webapp-testing (Anthropic Official Skill)

webapp-testing is an official Anthropic skill for testing local web applications with Playwright, and per the Zima Store's 2026 guide to AI agent skills, it helps Claude Code interact with a running app, capture screenshots, inspect console logs, debug UI behavior, and verify frontend functionality. Its value for beginners is the mindset shift it produces: as the Zima guide puts it, Claude Code may edit the correct files but still leave the app broken — a browser-testing skill pushes the workflow from 'I changed the code' to 'I opened the app and verified the behavior.' The skill is best suited to verifying login flows, forms, dashboard behavior, browser errors, and UI regressions on locally running applications. Because it comes from Anthropic's official skills repository, it carries stronger trust signals than most community alternatives — the Zima guide notes that official repos from Anthropic, Vercel, Supabase, and others are considered safe defaults, and Firecrawl's 2026 skills guide similarly recommends official and partner repositories over less-trusted community sources. For a beginner who wants their agent to actually confirm that changes work rather than just claim they do, this is the natural first official skill to install.

Best for: Beginners who want their coding agent to verify app behavior in a real browser, not just edit code and hope for the best

Pricing: Free (open source)

Key features

• Official Anthropic skill from the anthropics/skills repository
• Tests local web applications through Playwright in a real browser
• Captures screenshots during test runs
• Inspects browser console logs for errors
• Debugs UI behavior and verifies frontend functionality
• Suited to login flows, forms, dashboards, and UI regression checks
• Listed among the top 10 AI agent skills for Claude Code in 2026 by Zima Store

Pros

• Official Anthropic skill with strong trust signals
• Closes the gap between 'code changed' and 'feature verified'
• Screenshot and console-log inspection makes debugging accessible to beginners
• Free and open source

Cons

• Scoped to local web apps rather than full cross-browser or production test suites
• Playwright-based, so it overlaps with dedicated E2E skills — curate to avoid duplication
• Does not cover API, performance, or mobile testing

Visit webapp-testing (Anthropic Official Skill)

#5 vitest-unit-testing

Vitest has, according to QASkills.sh, eaten most of the unit-testing market from Jest by 2026, and the vitest-unit-testing skill is the natural entry point for beginners adding unit-test coverage to JavaScript and TypeScript projects. Per QASkills.sh's 2026 guide, the skill teaches Claude when to use describe.concurrent for parallel test execution, how to set up MSW (Mock Service Worker) for fetch mocking, and how to write deterministic snapshot tests that do not flake. Unit testing is the lowest-risk way for a beginner to start using agent skills: the blast radius is small, feedback is instant, and the agent's output is easy to review. QASkills.sh reports an average install time under two minutes and a quality score of 94 — second only to the Playwright E2E skill in its comparison table. In a realistic workflow described by QASkills.sh, a QA engineer asks Claude Code to review the week's merged pull requests and propose missing tests; with vitest-unit-testing loaded, Claude reads the diff and writes the missing unit tests itself, with the full Vitest suite then fanning out 12,000 tests in 90 seconds. That loop is exactly the kind of repeatable, high-leverage work skills were designed for.

Best for: JavaScript and TypeScript beginners who want fast, low-risk wins by having Claude generate reliable unit tests

Pricing: Free (install via npx @qaskills/cli add vitest-unit-testing --agent claude-code)

Key features

• Teaches Claude when to use describe.concurrent for parallel unit tests
• MSW (Mock Service Worker) setup for fetch mocking
• Deterministic snapshot test patterns that avoid flakiness
• Install time under 2 minutes per QASkills.sh
• Quality score of 94 in QASkills.sh's 2026 comparison
• Targets Vitest, the dominant JS/TS unit testing framework of 2026
• Works well in a diff-review workflow for proposing missing tests on merged PRs

Pros

• High quality score (94) and very fast install
• Encodes modern Vitest patterns like concurrency and MSW mocking
• Unit tests are easy for beginners to review and validate
• Pairs naturally with PR-review workflows

Cons

• Vitest-only — teams still on Jest need a different skill
• Does not cover integration or E2E testing
• Generated tests still need human review for meaningful assertions

Visit vitest-unit-testing

#6 pytest-patterns

For Python teams, QASkills.sh calls pytest-patterns the must-have testing skill of 2026, and it earns its place on this beginner list because pytest is usually the first framework new Python testers encounter. According to QASkills.sh's 2026 guide, the skill covers fixtures, parametrize, markers, and modern pytest-asyncio patterns, and it makes Claude notably good at structuring conftest.py hierarchies and avoiding fixture-scope mistakes — two of the most common ways beginners create slow, confusing test suites. Fixture scope errors in particular are a classic beginner trap: a wrongly scoped fixture can cause tests to share state silently or re-run expensive setup hundreds of times. By encoding correct scope and hierarchy conventions into the skill, the agent produces test suites that behave predictably from day one. QASkills.sh reports an average install time under two minutes and a quality score of 95, tying for the highest in its comparison table alongside the Playwright skill. The skill covers both unit and integration testing patterns, making it a two-for-one foundation for a Python QA stack. Combined with a code-review pass, it lets a beginner go from an empty tests folder to a well-structured, idiomatic pytest suite in a single working session.

Best for: Python beginners and QA engineers who want idiomatic, well-structured pytest suites without learning fixture pitfalls the hard way

Pricing: Free (install via npx @qaskills/cli add pytest-patterns --agent claude-code)

Key features

• Covers pytest fixtures, parametrize, and markers
• Modern pytest-asyncio patterns for async Python code
• Structures conftest.py hierarchies correctly
• Helps avoid fixture-scope mistakes that cause flaky or slow suites
• Covers both unit and integration testing
• Install time under 2 minutes per QASkills.sh
• Quality score of 95 in QASkills.sh's 2026 comparison table

Pros

• Tied for the highest quality score (95) in QASkills.sh's comparison
• Explicitly targets common beginner mistakes like fixture scope errors
• Covers both unit and integration testing in one skill
• Fast install and Python-native

Cons

• Python-only — no value for JavaScript or polyglot teams
• Assumes pytest as the framework; unittest users get less benefit
• Async patterns require some pytest-asyncio familiarity to review confidently

Visit pytest-patterns

#7 accessibility-axe-pa11y

Accessibility testing has moved from nice-to-have to default deliverable: according to QASkills.sh, WCAG 2.2 AA is a default deliverable in 2026, and the accessibility-axe-pa11y skill is the beginner-friendly way to bake it into an agent workflow. Per QASkills.sh's 2026 guide, the skill teaches Claude to run axe-core inside Playwright, parse the violations it finds, and convert each violation into a GitHub issue automatically — turning what used to be a separate audit engagement into a routine part of every test run. For beginners, accessibility is an area where agents genuinely help: axe-core produces dense, technical violation reports, and the skill teaches the agent to interpret them, prioritize fixes, and track them to closure rather than dumping raw output on the user. The skill pairs naturally with an E2E skill — QASkills.sh's example workflow has a QA engineer ask Claude to 'add axe-core to every new Playwright test' after the E2E suite is written, and the accessibility skill takes over from there. QASkills.sh reports an average install time under five minutes and a quality score of 92. For teams facing accessibility compliance requirements, this is one of the highest-ROI skills to install early.

Best for: Beginners and teams who need to meet WCAG 2.2 AA requirements and want accessibility checks built into every test run

Pricing: Free (install via npx @qaskills/cli add accessibility-axe-pa11y --agent claude-code)

Key features

• Runs axe-core accessibility checks inside Playwright tests
• Parses and interprets WCAG violations
• Converts each accessibility violation into a GitHub issue automatically
• Supports WCAG 2.2 AA, described by QASkills.sh as a 2026 default deliverable
• Pairs with E2E skills for a combined browser + a11y workflow
• Install time under 5 minutes per QASkills.sh
• Quality score of 92 in QASkills.sh's 2026 comparison

Pros

• Automates issue creation so violations get tracked, not just reported
• High quality score (92) and quick install
• Makes a complex compliance domain approachable for beginners
• Complements E2E skills cleanly

Cons

• Automated axe-core checks catch only a subset of WCAG issues; manual testing is still needed
• Requires an existing Playwright setup to be most effective
• GitHub-issue workflow assumes a GitHub-based team

Visit accessibility-axe-pa11y

#8 k6-load-testing

Performance testing intimidates most beginners, which is exactly why the k6-load-testing skill earns a spot here: according to QASkills.sh, it is the most-installed performance skill for Claude Code in 2026. Per QASkills.sh's 2026 guide, the skill covers k6 scripting, threshold definitions, and the four core load-test shapes — smoke, load, stress, and soak — so the agent can recommend and generate the right test profile for the question you are actually asking. It also covers parsing results from the newer k6 cloud SaaS offering, letting teams run heavy tests in the cloud and have Claude interpret the output. Load testing involves real infrastructure risk, so a skill that encodes correct ramp-up patterns, sensible thresholds, and the difference between a smoke test and a stress test protects beginners from both bad data and accidental self-inflicted outages. QASkills.sh reports an average install time under five minutes and a quality score of 90 in its comparison table. For a beginner, the practical workflow is simple: describe the endpoint and expected traffic, and the skill guides the agent to produce a k6 script with realistic user behavior, defined latency and error-rate thresholds, and a test shape appropriate to the goal.

Best for: Beginners who need to validate API and application performance without learning k6 scripting from scratch

Pricing: Free (install via npx @qaskills/cli add k6-load-testing --agent claude-code; k6 Cloud is a separate paid product)

Key features

• k6 scripting guidance for performance tests
• Threshold definitions for latency and error rates
• Covers smoke, load, stress, and soak test shapes
• Parses results from k6 cloud SaaS runs
• Most-installed performance skill per QASkills.sh
• Install time under 5 minutes per QASkills.sh
• Quality score of 90 in QASkills.sh's 2026 comparison table

Pros

• Most-installed performance skill for Claude Code in 2026
• Teaches the distinction between test shapes, preventing wrong-test mistakes
• Solid quality score (90)
• Works with both local k6 and k6 cloud results

Cons

• k6-specific; teams using Locust, JMeter, or Artillery need different tooling
• Load testing against shared environments carries risk and needs coordination
• k6 Cloud usage involves separate costs beyond the skill itself

Visit k6-load-testing

#9 ci-github-actions-test-matrix

The final piece of a beginner's QA skill stack is making everything run automatically, and ci-github-actions-test-matrix is the skill QASkills.sh describes as wrapping all the other testing skills into one GitHub Actions matrix. According to QASkills.sh's 2026 guide, the skill teaches Claude to build a GitHub Actions workflow that runs unit, API, E2E, and accessibility tests in parallel, with sensible concurrency limits and cache keys so runs stay fast and do not hammer GitHub's runners. For beginners, CI configuration is one of the most frustrating parts of testing: YAML is unforgiving, caching rules are obscure, and a misconfigured matrix can silently skip tests or burn through CI minutes. The skill encodes those conventions so the agent produces a working pipeline on the first attempt rather than after a week of trial and error. QASkills.sh reports an average install time under ten minutes and a quality score of 93 — third-highest in its comparison table. Their example workflow shows the payoff: after a QA engineer writes unit, Playwright, and accessibility tests with other skills, this skill wires all four job types into the workflow, producing a successful matrix run with 16 parallel shards all green.

Best for: Beginners who have written tests with other skills and now need a reliable, parallel CI pipeline to run them automatically

Pricing: Free (install via npx @qaskills/cli add ci-github-actions-test-matrix --agent claude-code; GitHub Actions usage limits apply per GitHub's plans)

Key features

• Builds GitHub Actions workflows that run unit, API, E2E, and accessibility tests in parallel
• Matrix-based test sharding across CI runners
• Sensible concurrency limits to avoid runner overload
• Cache-key configuration for faster runs
• Install time under 10 minutes per QASkills.sh
• Quality score of 93 in QASkills.sh's 2026 comparison table
• Acts as the glue skill connecting all other testing skills into CI

Pros

• Consolidates all testing types into one parallel workflow
• High quality score (93)
• Eliminates the most common beginner YAML and caching mistakes
• Works with free GitHub-hosted runners for public repos

Cons

• GitHub Actions-specific — teams on GitLab CI or CircleCI need different skills
• Should be installed last, after the test skills it wires together
• Private-repo CI minute limits may apply depending on your GitHub plan

Visit ci-github-actions-test-matrix

Conclusion

Testing and QA is one of the best domains for beginners to start using Claude skills in 2026: the feedback is concrete, the workflows are repeatable, and the skills above — from Playwright E2E and Vitest unit testing to accessibility checks, load testing, and CI matrices — map cleanly onto real QA jobs. But the ecosystem's biggest risk is also its biggest opportunity: with research suggesting roughly 26% of community skills carry at least one vulnerability, what you install matters as much as what you build. That is exactly why SkillzBench exists. SkillzBench is building a free, independent directory that verifies every skill with security scans, license checks, and version pinning, benchmarks every skill in a sandbox to publish real Lift numbers, and ships a free MCP connector with per-agent loadouts so your agent only ever sees the skills you chose. Everything — the directory, the MCP, benchmarks, and alerts — is free, funded by ads and never by users. Want skills, proven? Join the SkillzBench waitlist and newsletter today to be first in line when the directory, benchmarks, and free MCP connector go live, and start building a QA skill stack you can actually trust.

Frequently asked questions

What exactly is a Claude skill, and how is it different from a prompt?

A Claude skill is a folder containing a SKILL.md file — a versioned, validated, repeatable instruction set with optional bundled scripts and reference files. Unlike a prompt, which you type once and lose, a skill loads automatically when Claude Code detects a relevant task. Per Anthropic's docs, only the skill's name and description (roughly 100 tokens) are visible at startup, with full instructions loading on demand via progressive disclosure. Skills live in ~/.claude/skills (personal) or .claude/skills (project, shared via git).

Are Claude skills safe to install?

Not all of them. Published research cited by SkillzBench found that roughly 26% of agent skills carry at least one vulnerability. Community skills can include undocumented network calls or bundled scripts with side effects, so Firecrawl's 2026 guide recommends auditing every SKILL.md and bundled script and favoring official repositories like anthropics/skills. SkillzBench is building exactly this safety layer: every skill it lists is security-scanned, license-checked, validated, and pinned to a commit SHA, with a 0–100 trust score combining security, quality, license, maintenance, and community signals.

How many testing skills should a beginner install at once?

Fewer than you might think. QASkills.sh's 2026 guide warns that installing every popular skill at once is a common mistake: skills cost tokens, and with 40 skills installed Claude has to scan all of them per task. Their recommendation is to stick to the 8–12 skills that actually match your stack, and to avoid mixing overlapping skills — for example, do not install both a Playwright E2E skill and a Cypress E2E skill in the same repo, since Claude will pick whichever it sees first alphabetically.

How do I know if a skill actually improves my testing results?

Look for evidence, not marketing. SkillzBench's approach is to benchmark every skill in a sandbox with the same model, run with and without the skill, and publish the Lift in percentage points — so you can see whether a skill measurably improves outcomes rather than just trusting its self-description. Lift is always shown separately from ratings, and money never touches scores. Until SkillzBench's benchmarks go live, you can evaluate a skill yourself by checking whether its description reads like a clear routing rule, whether it includes concrete worked examples, and whether its bundled code actually delivers what it claims.

Do these testing skills cost anything?

The skills themselves in this list are free — most are open-source SKILL.md files installed with a single npx command, and SkillzBench's entire platform (directory, MCP connector, benchmarks, and alerts) is free, funded by advertising rather than user fees. Indirect costs can apply: skills consume model tokens when loaded, GitHub Actions usage counts against your plan's minutes for private repos, and managed services like k6 Cloud charge separately. No skill on this list requires a paid subscription.

Can I use these skills with agents other than Claude Code?

Often, yes. Skills that follow the Agent Skills Standard format — like the qa-skills 42-skill library highlighted in this list — work drop-in with Claude Code, OpenAI Codex, Cursor, Gemini CLI, and any compatible runtime. Skills-hub.ai also notes its SKILL.md skills install into Claude Code, Cursor, Codex CLI, Windsurf, or any MCP-compatible tool. SkillzBench is building a free MCP server with per-agent loadouts specifically so you can connect any agent to only the skills and tools you choose for that agent.

What is SkillzBench and when will it be available?

SkillzBench (skillzbench.com) is AIQ Labs' free, independent directory of AI agent skills plus a free MCP tool connecting any agent to the skills you choose. It is built around five pillars: Discover (crawling and deduplicating skills from GitHub), Verify (security scans, license checks, validation, and commit-SHA version pinning), Benchmark (sandbox testing that publishes Lift in percentage points), Equip (one free MCP server plus per-agent loadouts), and Stay current (tracking upstream changes with re-scans, re-benchmarks, and alerts). SkillzBench is currently pre-launch and actively being built — the waitlist and newsletter are live now, and joining is the best way to get access as soon as the directory, benchmarks, and MCP connector become available.