SkillProof tested Claude skills

Best Claude Skills for Testing (2026)

Most "best skills" lists are copied README summaries. This one isn't: we installed all 144 testing skills in our catalog, ran each against a baseline and scored four criteria. 49 passed cleanly, average score 8.6/10. The one to start with: Screen Reader Testing.

Ranked by SkillProof Score · how we score

  1. 1

    Screen Reader Testing

    Screen reader testing playbook for VoiceOver, NVDA, and JAWS with ARIA fixes.

    From our test: Given a modal with an icon-only close button and no dialog role, it produced the exact aria-label and role=dialog fixes plus concrete VoiceOver and NVDA test scripts a generic answer skipped entirely.

    Tested · Works ★ 9.6/10 by wshobson

  2. 2

    Skill Security Auditor

    OWASP-mapped code/secrets/config audit with working bundled scan scripts

    From our test: Run against a 13-line file with SQL injection, command injection and a hardcoded API key, its scripts caught all three plus a missing .gitignore issue that a manual review missed.

    Tested · Works ★ 9.6/10 by eigent-ai

  3. 3

    Web Quality Audit

    Lighthouse-style audit across Performance, Accessibility, SEO, and Best Practices with severity tiers.

    From our test: Auditing a deliberately broken test page, it caught the charset and mixed-content issues a flat review missed, tagging every finding with file:line and a concrete fix instead of a bare list.

    Tested · Works ★ 9.6/10 by addyosmani

  4. 4

    OSINT Methodology

    Structured 5-stage external recon methodology with confidence levels, severity rubric, and OpSec rules.

    From our test: The companion `offensive-osint` skill it repeatedly points to for implementation detail actually exists in the same repo and loads clean; an authorized-recon mini-task run through it produced…

    Tested · Works ★ 9.6/10 by elementalsouls

  5. 5

    Jadx

    APK decompile + a systematic grep checklist for secrets, weak crypto, SQLi, and WebView holes.

    From our test: Planted 6 realistic vulns (hardcoded API key/password, MD5 hashing, string-concat SQL rawQuery, JS-enabled WebView with addJavascriptInterface, MODE_WORLD_READABLE prefs) in a fake decompiled tree…

    Works with setup ★ 9.6/10 by BrownFineSecurity

  6. 6

    Jtagprobe

    SWD/JTAG debug-port pentest probe via a physical SEGGER J-Link, classifies OPEN/LOCKED/DEAD.

    From our test: This is a real 195-line Python tool (not vaporware) wired to SEGGER's JLinkExe; fed it the sample DPIDR/all-0xFF readout from its own docs and it correctly reasoned LOCKED-not-erased (a successful…

    Tested · Works ★ 9.6/10 by BrownFineSecurity

  7. 7

    Swift Testing

    Swift Testing patterns, test-double taxonomy, and migration guidance for @Test-based suites.

    From our test: Asked for a test suite for an email validator plus a repository mock: the skill version used @Suite/@Test parameterization across 3-4 inputs each, correctly renamed the naive 'MockUserRepository' to…

    Tested · Works ★ 9.6/10 by bocato

  8. 8

    Arize Dataset

    CLI-driven Arize dataset CRUD, versioning, and export for LLM eval sets.

    From our test: Produced exact `ax datasets create --file -` / `ax datasets append --json` syntax straight from the docs (right subcommand plurality, positional NAME_OR_ID, --space requirement) where a cold guess…

    Tested · Works ★ 9.6/10 by github

  9. 9

    SkillHone Synthesis

    Turns tool exploration into hard, machine-gradable benchmark Q/A instead of trivia lookups.

    From our test: Built a 6-employee toy dataset and compared a naive question ('what's E003's salary?', directly readable) against a skill-guided one ('among Engineering hires before 2022, excluding contractors, who…

    Tested · Works ★ 9.6/10 by Tencent

  10. 10

    Cypress Author

    Writes, updates, and fixes Cypress E2E and component tests following house style rules.

    From our test: Same login-test request produced a hardcoded URL and a bare cy.wait(2000) without the skill; following its author-rules.md swapped in a relative baseUrl visit, data-cy selectors, and a timeout-based…

    Tested · Works ★ 9.6/10 by cypress-io

How this ranking works

Each skill gets a fresh install in Claude Code, three trigger phrasings, and the same task run with and without it. Scores: installation (5), trigger reliability (5), output vs baseline (10), documentation honesty (5) — normalized to a 10-point SkillProof Score. Full rubric on the methodology page. Queue-verdict skills sit at the bottom until testing completes.

New here? Start with how to install Claude skills — or grab the whole setup in one command with a curated pack ($10).

On a different AI tool? These tested instructions adapt: see our conversion guides for Cursor rules, Copilot instructions, Windsurf and Custom GPTs.