Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →(Proposal, unverified) Judge whether argumentative relations between unit-classification's rhetorical labels actually hold in a paper (e.g. is an AIM label adequately substantiated by BACKGROUND labels) — a second-order quality judgment over already-classified units, not raw text. Use this after unit-classification has labeled a paper's units with a rhetorical/argumentative label set, when the user wants to know if the paper's argument structure is actually sound, not just what role each sentenc
.claude/skills/yogsoth-ai-rhetorical-structure-quality/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -25% | 0% |
Second-order SOP: judges whether rhetorical/argumentative labels (from unit-classification) actually substantiate each other, e.g. AIM vs BACKGROUND. Fills a gap in the evaluative-stance × content-layer matrix (quality-judgment × argumentative-rhetorical-role) that no verified method covers — this is a design proposal, not a transcription of an established methodology.
Subagent — spawned via spawn-agent skill.
This SOP has no primary-source precedent (unlike CoreSC/AZ, which it consumes labels from). Its description explicitly says "(Proposal, unverified)" so it is never triggered with the same implied confidence as a verified method. Do not remove that qualifier from the description without re-validating the method against real usage first.
<!-- BEGIN available-tables (generated) -->
| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,287 | 14,486 | +9% | 1 | 1 | 0% | 1,276 | 1,845 | +45% | 0 | 0 | — |
case-02 | fail→fail | 15,308 | 27,628 | +80% | 1 | 1 | 0% | 1,496 | 1,520 | +2% | 0 | 0 | — |
case-03 | fail→fail | 14,421 | 32,414 | +125% | 1 | 1 | 0% | 1,339 | 1,154 | -14% | 0 | 0 | — |
case-04 | fail→fail | 10,783 | 14,559 | +35% | 1 | 1 | 0% | 935 | 1,891 | +102% | 0 | 0 | — |
case-05 | fail→fail | 6,687 | 12,094 | +81% | 1 | 1 | 0% | 251 | 1,381 | +450% | 0 | 0 | — |
case-06 | pass→pass | 14,418 | 13,099 | -9% | 1 | 1 | 0% | 1,531 | 1,517 | -1% | 0 | 0 | — |
case-12 | fail→pass | 23,661 | 11,799 | -50% | 1 | 1 | 0% | 2,641 | 1,395 | -47% | 0 | 0 | — |
case-07 | pass→pass | 6,450 | 6,512 | +1% | 1 | 1 | 0% | 226 | 431 | +91% | 0 | 0 | — |
case-08 | fail→fail | 9,421 | 21,948 | +133% | 1 | 1 | 0% | 651 | 3,217 | +394% | 0 | 0 | — |
case-09 | fail→pass | 21,620 | 17,368 | -20% | 1 | 1 | 0% | 2,989 | 2,337 | -22% | 0 | 0 | — |
case-10 | fail→fail | 35,995 | 49,461 | +37% | 1 | 1 | 0% | 5,204 | 8,479 | +63% | 0 | 0 | — |
case-11 | fail→pass | 19,349 | 10,995 | -43% | 1 | 1 | 0% | 2,567 | 1,181 | -54% | 0 | 0 | — |
case-13 | fail→fail | 13,888 | 13,882 | -0% | 1 | 1 | 0% | 1,340 | 1,546 | +15% | 0 | 0 | — |
case-14 | fail→fail | 12,827 | 17,629 | +37% | 1 | 1 | 0% | 1,241 | 767 | -38% | 0 | 0 | — |
case-15 | fail→pass | 20,698 | 17,752 | -14% | 1 | 1 | 0% | 2,398 | 2,327 | -3% | 0 | 0 | — |
case-16 | fail→pass | 21,678 | 14,652 | -32% | 1 | 1 | 0% | 2,473 | 1,855 | -25% | 0 | 0 | — |
case-17 | fail→fail | 25,680 | 23,944 | -7% | 1 | 1 | 0% | 3,227 | 3,255 | +1% | 0 | 0 | — |
case-18 | fail→fail | 16,991 | 19,733 | +16% | 1 | 1 | 0% | 1,888 | 2,423 | +28% | 0 | 0 | — |
case-19 | fail→pass | 22,819 | 9,227 | -60% | 1 | 1 | 0% | 2,837 | 904 | -68% | 0 | 0 | — |
case-20 | fail→fail | 11,050 | 27,384 | +148% | 1 | 1 | 0% | 977 | 3,997 | +309% | 0 | 0 | — |
case-21 | pass→pass | 17,964 | 9,792 | -45% | 1 | 1 | 0% | 2,021 | 1,060 | -48% | 0 | 0 | — |
case-22 | fail→fail | 26,934 | 22,720 | -16% | 1 | 1 | 0% | 3,471 | 2,990 | -14% | 0 | 0 | — |
case-23 | fail→pass | 17,400 | 11,204 | -36% | 1 | 1 | 0% | 1,947 | 1,298 | -33% | 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 21 counted toward the lift figure. The other 2 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 +30 percentage points is the difference between those two pass rates over the 21 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.