Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply a team's writing style consistently — extract the house style from exemplar documents into a checkable rule card, run the conformance pass on new drafts, and fix violations without flattening the author's voice. Use when asked make this match our style, why do our docs all sound different, build a style guide from our best docs, or check this draft against house style. Produces the extracted rule card, the conformance pass with per-fix reasons, and the voice-preservation line.
.claude/skills/mohitagw15856-house-style-enforcer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 271% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 120% | 0% |
Most teams have a house style nobody wrote down — it lives in the three documents everyone calls "really good" and gets enforced by vibes in review. This skill makes it checkable: extract the rules from the exemplars (structure, tone registers, formatting conventions, banned constructions), compress them into a one-page rule card, then run drafts against the card — fixing violations of the card while leaving the author's voice alone, because style guides that flatten everyone into one android author get ignored in self-defense.
Ask for these if not provided:
Structure: rules + exemplar cites] · Register: ...] · Formatting: ...] · Settled fights: the one-liners] The voice line: this card does not govern vocabulary within register, rhythm, personality].
| Location | Violation | Card rule | The fix | |---|---|---|---| Plus: "flagged but not card-covered: items] — propose as rules or leave alone"]
Add-one-consider-killing-one · card owner · where it lives]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 16,549 | 16,560 | +0% | 1 | 1 | 0% | 2,657 | 3,882 | +46% | 0 | 0 | — |
case-02 | fail→pass | 17,532 | 14,318 | -18% | 1 | 1 | 0% | 3,890 | 4,129 | +6% | 0 | 0 | — |
case-03 | fail→pass | 4,706 | 10,894 | +131% | 1 | 1 | 0% | 790 | 2,932 | +271% | 0 | 0 | — |
case-04 | fail→pass | 34,751 | 15,286 | -56% | 1 | 1 | 0% | 6,175 | 3,471 | -44% | 0 | 0 | — |
case-05 | fail→pass | 5,553 | 6,938 | +25% | 1 | 1 | 0% | 993 | 2,181 | +120% | 0 | 0 | — |
case-06 | fail→fail | 12,046 | 15,080 | +25% | 1 | 1 | 0% | 2,075 | 3,698 | +78% | 0 | 0 | — |
case-07 | pass→pass | 9,878 | 8,931 | -10% | 1 | 1 | 0% | 1,936 | 2,512 | +30% | 0 | 0 | — |
case-08 | fail→fail | 2,325 | 5,698 | +145% | 1 | 1 | 0% | 440 | 2,191 | +398% | 0 | 0 | — |
case-09 | fail→pass | 10,682 | 7,076 | -34% | 1 | 1 | 0% | 1,649 | 2,383 | +45% | 0 | 0 | — |
case-10 | fail→pass | 9,478 | 11,773 | +24% | 1 | 1 | 0% | 1,866 | 3,313 | +78% | 0 | 0 | — |
case-11 | fail→pass | 10,876 | 7,899 | -27% | 1 | 1 | 0% | 1,714 | 2,377 | +39% | 0 | 0 | — |
case-12 | fail→fail | 12,769 | 13,162 | +3% | 1 | 1 | 0% | 2,081 | 3,102 | +49% | 0 | 0 | — |
case-13 | pass→pass | 7,267 | 6,777 | -7% | 1 | 1 | 0% | 1,314 | 2,340 | +78% | 0 | 0 | — |
case-14 | fail→pass | 12,055 | 4,640 | -62% | 1 | 1 | 0% | 1,793 | 1,820 | +2% | 0 | 0 | — |
case-15 | pass→pass | 12,521 | 14,277 | +14% | 1 | 1 | 0% | 2,047 | 3,063 | +50% | 0 | 0 | — |
case-16 | pass→pass | 11,932 | 12,024 | +1% | 1 | 1 | 0% | 1,932 | 3,069 | +59% | 0 | 0 | — |
case-17 | pass→pass | 7,537 | 5,176 | -31% | 1 | 1 | 0% | 1,301 | 2,090 | +61% | 0 | 0 | — |
case-18 | pass→pass | 11,023 | 8,695 | -21% | 1 | 1 | 0% | 1,794 | 2,450 | +37% | 0 | 0 | — |
case-19 | fail→fail | 6,171 | 5,232 | -15% | 1 | 1 | 0% | 1,135 | 1,960 | +73% | 0 | 0 | — |
case-20 | pass→pass | 11,503 | 9,502 | -17% | 1 | 1 | 0% | 1,756 | 2,699 | +54% | 0 | 0 | — |
case-21 | fail→pass | 8,960 | 6,205 | -31% | 1 | 1 | 0% | 1,457 | 2,011 | +38% | 0 | 0 | — |
case-22 | fail→pass | 8,184 | 4,031 | -51% | 1 | 1 | 0% | 1,484 | 1,804 | +22% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +50 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.