Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use only when the user explicitly names crossframe-critical for a Chinese structural critique dossier, article plan, or long-form critical essay.
.claude/skills/sickn33-crossframe-critical/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 88% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 180% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 26% | 0% |
crossframe-critical, $crossframe-critical, or asks to test this critical parallel skill.crossframe-suite route.This AAS-ready copy preserves the original CrossFrame skill body below. Chinese remains the canonical semantic layer; English metadata is only for discovery, installation, and repository review.
crossframe-suite routing; do not apply it as a generic default reasoning layer.> 本 skill 不独立触发。 所有 CrossFrame 任务统一从 crossframe-suite 入口调度。用户无需直接调用本 skill;suite 根据路由规则在需要时自动加载。
This is a parallel local test skill. It does not replace crossframe, crossframe-essay, crossframe-public, or crossframe-suite.
crossframe-critical writes critical Chinese essays that first use CrossFrame to establish structure, evidence boundaries, scale, mechanism candidates, and judgment grade, then sharpen the output into critique.
The critique may absorb Marxist problem awareness: interests, cost transfer, alienation, commodification, ideology, naturalized domination, and reproduction of conditions. It must not mechanically force every topic into class/capital language.
On every trigger, read:
../crossframe/SKILL.md../crossframe/references/read-routing-map.md../crossframe/templates/read-state-capsule.md as v5-read-state-capsule and run ../crossframe/worksheets/source-anchor-integrity-check.md; if the capsule is missing, return to ../crossframe/SKILL.md instead of inventing source routing here.protocols/critical-article-protocol.mdreferences/critical-matrix.mdreferences/example-and-evidence-rules.md../crossframe/references/source-ledger-workflow.md 并建立来源台账。templates/critical-output-template.mdIf the topic needs long-form style control, also read ../crossframe-essay/SKILL.md and reuse only its article discipline, not its whole output contract.
Default output has exactly three visible sections:
text# 批判底稿 # 篇章方案 # 正文
Do not collapse the result into a short answer, checklist, memo, or diagnosis summary unless the user explicitly asks for that.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 44,436 | 30,789 | -31% | 1 | 1 | 0% | 6,031 | 5,091 | -16% | 0 | 0 | — |
case-02 | fail→pass | 41,353 | 36,778 | -11% | 1 | 1 | 0% | 5,957 | 5,977 | +0% | 0 | 0 | — |
case-03 | pass→pass | 25,823 | 42,065 | +63% | 1 | 1 | 0% | 4,069 | 7,065 | +74% | 0 | 0 | — |
case-04 | fail→fail | 15,114 | 35,169 | +133% | 1 | 1 | 0% | 2,034 | 6,455 | +217% | 0 | 0 | — |
case-05 | pass→fail | 8,606 | 36,058 | +319% | 1 | 1 | 0% | 2,002 | 6,935 | +246% | 0 | 0 | — |
case-06 | fail→fail | 17,193 | 36,058 | +110% | 1 | 1 | 0% | 3,010 | 5,627 | +87% | 0 | 0 | — |
case-07 | pass→pass | 28,962 | 30,610 | +6% | 1 | 1 | 0% | 3,866 | 5,812 | +50% | 0 | 0 | — |
case-08 | fail→pass | 14,488 | 18,419 | +27% | 1 | 1 | 0% | 2,141 | 4,028 | +88% | 0 | 0 | — |
case-09 | pass→pass | 18,452 | 37,473 | +103% | 1 | 1 | 0% | 2,468 | 6,200 | +151% | 0 | 0 | — |
case-10 | pass→pass | 19,638 | 27,985 | +43% | 1 | 1 | 0% | 2,724 | 5,138 | +89% | 0 | 0 | — |
case-11 | pass→pass | 20,156 | 32,173 | +60% | 1 | 1 | 0% | 2,727 | 5,093 | +87% | 0 | 0 | — |
case-12 | fail→pass | 17,578 | 38,421 | +119% | 1 | 1 | 0% | 2,279 | 6,390 | +180% | 0 | 0 | — |
case-13 | pass→pass | 23,478 | 41,458 | +77% | 1 | 1 | 0% | 3,170 | 6,508 | +105% | 0 | 0 | — |
case-14 | fail→pass | 23,769 | 38,743 | +63% | 1 | 1 | 0% | 3,277 | 5,741 | +75% | 0 | 0 | — |
case-15 | fail→pass | 30,494 | 32,440 | +6% | 1 | 1 | 0% | 4,286 | 5,393 | +26% | 0 | 0 | — |
case-16 | fail→pass | 16,557 | 32,856 | +98% | 1 | 1 | 0% | 1,747 | 5,697 | +226% | 0 | 0 | — |
case-17 | pass→pass | 19,639 | 30,674 | +56% | 1 | 1 | 0% | 3,191 | 5,684 | +78% | 0 | 0 | — |
case-18 | pass→pass | 17,596 | 40,222 | +129% | 1 | 1 | 0% | 2,496 | 6,289 | +152% | 0 | 0 | — |
case-19 | pass→pass | 22,509 | 40,202 | +79% | 1 | 1 | 0% | 1,819 | 7,213 | +297% | 0 | 0 | — |
case-20 | pass→pass | 19,849 | 33,585 | +69% | 1 | 1 | 0% | 3,150 | 5,434 | +73% | 0 | 0 | — |
case-21 | fail→pass | 17,896 | 36,225 | +102% | 1 | 1 | 0% | 2,900 | 6,395 | +121% | 0 | 0 | — |
case-22 | fail→fail | 17,712 | 32,269 | +82% | 1 | 1 | 0% | 2,949 | 5,140 | +74% | 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 +27 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.