Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Reads the target file, analyzes it against Signal Theory principles, and suggests or applies edits to maximize signal-to-noise ratio. Can focus on specific dimensions: clarity (remove ambiguity), accuracy (fact-check against knowledge base), conciseness (cut noise), or tone (match receiver).
.claude/skills/miosa-osa-edit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -77% | 0% |
> Revise existing content for clarity, accuracy, and signal-to-noise ratio.
/edit <file> [--focus <clarity|accuracy|conciseness|tone>] [--for <person>]Reads the target file, analyzes it against Signal Theory principles, and suggests or applies edits to maximize signal-to-noise ratio. Can focus on specific dimensions: clarity (remove ambiguity), accuracy (fact-check against knowledge base), conciseness (cut noise), or tone (match receiver).
bash# Edit for overall quality /edit docs/pitch-card.md # Focus on conciseness /edit docs/architecture/FULL-SYSTEM-ARCHITECTURE.md --focus conciseness # Edit for a specific receiver /edit outreach-email.md --for "Ed Honour"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | pass→pass | 42,114 | 17,642 | -58% | 1 | 1 | 0% | 8,226 | 3,727 | -55% | 0 | 0 | — |
case-01 | fail→fail | 6,003 | 5,625 | -6% | 1 | 1 | 0% | 953 | 538 | -44% | 0 | 0 | — |
case-02 | fail→fail | 3,484 | 4,583 | +32% | 1 | 1 | 0% | 530 | 536 | +1% | 0 | 0 | — |
case-03 | pass→pass | 6,980 | 2,508 | -64% | 1 | 1 | 0% | 1,223 | 738 | -40% | 0 | 0 | — |
case-04 | pass→pass | 8,480 | 3,237 | -62% | 1 | 1 | 0% | 1,396 | 696 | -50% | 0 | 0 | — |
case-05 | fail→pass | 7,507 | 3,159 | -58% | 1 | 1 | 0% | 1,215 | 603 | -50% | 0 | 0 | — |
case-06 | fail→pass | 9,923 | 3,562 | -64% | 1 | 1 | 0% | 1,680 | 953 | -43% | 0 | 0 | — |
case-07 | fail→pass | 11,572 | 10,237 | -12% | 1 | 1 | 0% | 1,760 | 1,616 | -8% | 0 | 0 | — |
case-08 | fail→fail | 11,980 | 8,265 | -31% | 1 | 1 | 0% | 2,040 | 1,627 | -20% | 0 | 0 | — |
case-09 | pass→pass | 10,668 | 6,820 | -36% | 1 | 1 | 0% | 1,820 | 1,422 | -22% | 0 | 0 | — |
case-10 | fail→pass | 6,830 | 2,019 | -70% | 1 | 1 | 0% | 1,089 | 609 | -44% | 0 | 0 | — |
case-12 | fail→fail | 2,606 | 5,766 | +121% | 1 | 1 | 0% | 417 | 1,172 | +181% | 0 | 0 | — |
case-13 | fail→fail | 4,617 | 6,862 | +49% | 1 | 1 | 0% | 724 | 1,395 | +93% | 0 | 0 | — |
case-14 | fail→pass | 12,997 | 1,508 | -88% | 1 | 1 | 0% | 2,238 | 509 | -77% | 0 | 0 | — |
case-15 | fail→pass | 6,942 | 1,744 | -75% | 1 | 1 | 0% | 1,109 | 509 | -54% | 0 | 0 | — |
case-16 | fail→pass | 10,430 | 1,595 | -85% | 1 | 1 | 0% | 1,880 | 525 | -72% | 0 | 0 | — |
case-17 | pass→pass | 7,046 | 6,036 | -14% | 1 | 1 | 0% | 1,270 | 1,324 | +4% | 0 | 0 | — |
case-18 | pass→pass | 6,812 | 4,776 | -30% | 1 | 1 | 0% | 1,018 | 1,010 | -1% | 0 | 0 | — |
case-19 | fail→pass | 6,077 | 2,587 | -57% | 1 | 1 | 0% | 890 | 711 | -20% | 0 | 0 | — |
case-20 | fail→pass | 7,679 | 2,309 | -70% | 1 | 1 | 0% | 983 | 639 | -35% | 0 | 0 | — |
case-21 | fail→pass | 18,118 | 2,231 | -88% | 1 | 1 | 0% | 2,911 | 642 | -78% | 0 | 0 | — |
case-22 | fail→pass | 4,964 | 5,931 | +19% | 1 | 1 | 0% | 711 | 1,226 | +72% | 0 | 0 | — |
case-23 | fail→pass | 8,257 | 2,163 | -74% | 1 | 1 | 0% | 1,275 | 603 | -53% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +52 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.