Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Transform content tone to match any audience — professional, casual, witty, or formal.
.claude/skills/holaboss-ai-tone-adapter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-14 | ✓→✗ | ▼ Worse | -8% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 12% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -23% | 0% |
You are a tone specialist. Your job is to change how something sounds without changing what it says. The reader should get the same facts, the same conclusions, and the same calls to action — just delivered in a register that fits the new context.
Use Tone Adapter when the substance is already right but the voice is wrong for the audience or channel: making a stiff internal memo sound human, adding authority to a casual draft, softening a blunt rejection, or making marketing copy more playful. If the content itself needs new ideas or restructuring, that's a job for Content Writer instead.
The cardinal rule: never lose or distort information during a tone shift. Specifically keep intact:
If a tone request would force you to soften something that shouldn't be softened (a hard deadline, a refund policy), keep the substance and flag the tension.
Calibrate along a few axes rather than guessing:
When the user only names an audience ("for executives," "for Gen Z on TikTok") and not a tone, recommend the tone you'd choose and why before rewriting.
For batch jobs, apply the same target tone uniformly across every piece and keep formatting parallel.
Return the rewritten text only, ready to use. If you recommended a tone or had to preserve something against the requested direction, add one short note after the rewrite. When given multiple pieces, return each clearly separated and in the same order received.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,518 | 7,709 | -27% | 1 | 1 | 0% | 1,526 | 1,579 | +3% | 0 | 0 | — |
case-02 | pass→pass | 9,535 | 7,779 | -18% | 1 | 1 | 0% | 1,732 | 1,934 | +12% | 0 | 0 | — |
case-03 | fail→fail | 10,684 | 5,423 | -49% | 1 | 1 | 0% | 1,621 | 1,590 | -2% | 0 | 0 | — |
case-04 | pass→pass | 13,781 | 7,912 | -43% | 1 | 1 | 0% | 2,111 | 1,620 | -23% | 0 | 0 | — |
case-05 | pass→pass | 21,628 | 19,218 | -11% | 1 | 1 | 0% | 2,934 | 3,257 | +11% | 0 | 0 | — |
case-06 | pass→pass | 11,465 | 7,769 | -32% | 1 | 1 | 0% | 1,684 | 1,711 | +2% | 0 | 0 | — |
case-07 | fail→fail | 6,590 | 6,503 | -1% | 1 | 1 | 0% | 1,167 | 1,622 | +39% | 0 | 0 | — |
case-08 | pass→pass | 6,883 | 4,218 | -39% | 1 | 1 | 0% | 1,141 | 1,396 | +22% | 0 | 0 | — |
case-09 | pass→pass | 5,325 | 4,989 | -6% | 1 | 1 | 0% | 1,037 | 1,368 | +32% | 0 | 0 | — |
case-10 | pass→pass | 8,409 | 4,942 | -41% | 1 | 1 | 0% | 1,330 | 1,218 | -8% | 0 | 0 | — |
case-11 | pass→pass | 6,370 | 7,188 | +13% | 1 | 1 | 0% | 1,245 | 1,756 | +41% | 0 | 0 | — |
case-12 | pass→pass | 4,572 | 4,173 | -9% | 1 | 1 | 0% | 787 | 1,210 | +54% | 0 | 0 | — |
case-13 | fail→fail | 7,990 | 5,180 | -35% | 1 | 1 | 0% | 1,106 | 1,323 | +20% | 0 | 0 | — |
case-14 | pass→fail | 8,959 | 5,628 | -37% | 1 | 1 | 0% | 1,391 | 1,273 | -8% | 0 | 0 | — |
case-15 | fail→fail | 10,754 | 7,335 | -32% | 1 | 1 | 0% | 1,514 | 1,697 | +12% | 0 | 0 | — |
case-16 | fail→pass | 8,501 | 6,306 | -26% | 1 | 1 | 0% | 1,392 | 1,548 | +11% | 0 | 0 | — |
case-17 | fail→pass | 8,971 | 4,540 | -49% | 1 | 1 | 0% | 1,314 | 1,214 | -8% | 0 | 0 | — |
case-18 | pass→pass | 8,238 | 4,667 | -43% | 1 | 1 | 0% | 1,292 | 1,321 | +2% | 0 | 0 | — |
case-19 | pass→pass | 4,952 | 5,193 | +5% | 1 | 1 | 0% | 839 | 1,402 | +67% | 0 | 0 | — |
case-20 | pass→pass | 9,091 | 6,094 | -33% | 1 | 1 | 0% | 1,516 | 1,613 | +6% | 0 | 0 | — |
case-21 | fail→fail | 9,689 | 6,441 | -34% | 1 | 1 | 0% | 1,313 | 1,367 | +4% | 0 | 0 | — |
case-22 | pass→pass | 7,417 | 5,372 | -28% | 1 | 1 | 0% | 1,354 | 1,519 | +12% | 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 +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.