Skills built and bench-tested by SkillProof — the tested Claude skills directory. We test other people's skills for a living (public methodology); ours go through the same protocol, and every skill ships with a measured benchmark, not marketing claims.
Each skill lives in its own repo — one git clone installs it.
| Skill | What it does | Measured result | Install |
|---|---|---|---|
| seo-translator | SEO-aware, human-sounding translation: per-market registers, never-translate list, anti-AI pass | 76k tokens / 4m53s per language (2,900-word article) | git clone https://github.com/Skillproofdev/seo-translator ~/.claude/skills/seo-translator |
| token-discipline | Input-side token savings: search-before-read, batched tool calls, no redundant re-reads — 9 enforceable rules | −19–20% tokens on multi-step work (5 controlled task pairs); a wash on one-shots | git clone https://github.com/Skillproofdev/token-discipline ~/.claude/skills/token-discipline |
| research-discipline | Verifiable research: live sources over memory, per-claim citations, independent 2-source check, [unverified] flags, adversarial pass — 8 enforceable rules | −59% wrong claims (13.0%→5.4%, 110 claims judged against live primary sources) at ~+10% tokens | git clone https://github.com/Skillproofdev/research-discipline ~/.claude/skills/research-discipline |
| review-discipline | Code review that proves every finding (demonstrated failure path or [suspicion]), breadth-before-depth | 18.5/21 seeded bugs caught vs 17 baseline, 0 false positives vs 1, ~100% demonstrated failure paths | git clone https://github.com/Skillproofdev/review-discipline ~/.claude/skills/review-discipline |
| email-discipline | Emails people answer: one goal, subject-as-ask, word budgets, no AI fluff | 'would reply now' 9/12 vs 3/12 baseline; mechanical pass 12/12 vs 2/12 | git clone https://github.com/Skillproofdev/email-discipline ~/.claude/skills/email-discipline |
| commit-discipline | Commit messages that match the diff: Conventional Commits, why-not-what, breaking footers | diff coverage 97% vs 83%, 0 fabrications vs 1, split-detect 2/2 vs 0/2 | git clone https://github.com/Skillproofdev/commit-discipline ~/.claude/skills/commit-discipline |
| api-discipline | REST/OpenAPI design that stays validator-clean and audits its own consistency | 0 validator errors vs 18 baseline, 0 consistency violations vs 6 | git clone https://github.com/Skillproofdev/api-discipline ~/.claude/skills/api-discipline |
| summary-discipline | Faithful summaries: numbers verbatim, contradictions flagged not silently resolved | 3/3 planted contradictions caught vs 0/3; 2 critical omissions vs 10 | git clone https://github.com/Skillproofdev/summary-discipline ~/.claude/skills/summary-discipline |
| readme-discipline | READMEs grounded in the code: every claim traces to a file, every command runs | section completeness 18/18 vs 14, blind preference 3/3 repos, trust 4.67 vs 3.67 | git clone https://github.com/Skillproofdev/readme-discipline ~/.claude/skills/readme-discipline |
| changelog-discipline | Changelogs traceable to real commits: nothing invented, breaking changes kept | commit traceability 87% vs 6%, format 30/36 vs 13/36 (fabrication a tie — both honest) | git clone https://github.com/Skillproofdev/changelog-discipline ~/.claude/skills/changelog-discipline |
| prompt-discipline | Prompt improvement that diagnoses before rewriting: minimal-diff + A/B test plan | task success 0.90 vs 0.70 original, half the prompt bloat (n=1, honest caveat) | git clone https://github.com/Skillproofdev/prompt-discipline ~/.claude/skills/prompt-discipline |
| text-humanizer | Remove AI tells while freezing every fact: rebuild the prose in a human voice, claim set immutable | beat the raw baseline on blind preference 8–4 (published 3 honest failures first); tells 2 vs 4, facts 129/130 | git clone https://github.com/Skillproofdev/text-humanizer ~/.claude/skills/text-humanizer |
SKILL.md validator, token calculator, Rules⇄SKILL.md converter — all client-side, no signup.
MIT.