Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →将中文毕业论文已有材料映射到“毕业论文角色”：把论文、模板、Overleaf 源稿、PDF、图表和实验材料按章节角色、研究问题和证据用途重新归位。 **Trigger**: 毕业论文材料映射, source role map, paper to chapter, 章节角色映射, 论文归章, 材料归位. **Use when**: 已经完成材料盘点，需要决定各份材料在毕业论文里扮演什么角色，而不是继续按原 paper 叙事直接写。 **Skip if**: 当前只是在修一个已经稳定的单章措辞，且没有新的来源材料进入。 **Network**: none. **Guardrail**: 不是做翻译；不是简单 `paper -> chapter` 分桶；必须显式说明“在毕业论文里的角色”。
.claude/skills/willoscar-thesis-source-role-mapper/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -3% | 0% |
这个 skill 负责把已有材料映射为毕业论文内部的角色与职责。
codex_md/material_index.mdcodex_md/question_list.mdcodex_md/chapter_role_map.mdAlways read:
references/overview.mdreferences/role_taxonomy.mdMachine-readable contract:
assets/source_role_map_contract.json至少要覆盖以下字段：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | pass→pass | 12,312 | 10,038 | -18% | 1 | 1 | 0% | 1,671 | 1,691 | +1% | 0 | 0 | — |
case-16 | pass→pass | 11,514 | 5,869 | -49% | 1 | 1 | 0% | 1,746 | 1,219 | -30% | 0 | 0 | — |
case-07 | fail→pass | 12,727 | 4,909 | -61% | 1 | 1 | 0% | 1,675 | 1,039 | -38% | 0 | 0 | — |
case-17 | fail→pass | 27,345 | 1,836 | -93% | 1 | 1 | 0% | 1,108 | 537 | -52% | 0 | 0 | — |
case-18 | pass→pass | 12,807 | 13,229 | +3% | 1 | 1 | 0% | 1,844 | 2,242 | +22% | 0 | 0 | — |
case-08 | fail→pass | 14,503 | 11,748 | -19% | 1 | 1 | 0% | 2,074 | 1,852 | -11% | 0 | 0 | — |
case-01 | fail→pass | 17,544 | 12,749 | -27% | 1 | 1 | 0% | 2,543 | 2,241 | -12% | 0 | 0 | — |
case-02 | pass→pass | 13,706 | 13,244 | -3% | 1 | 1 | 0% | 1,998 | 2,104 | +5% | 0 | 0 | — |
case-14 | pass→pass | 8,753 | 8,637 | -1% | 1 | 1 | 0% | 1,335 | 1,649 | +24% | 0 | 0 | — |
case-03 | fail→pass | 14,977 | 12,381 | -17% | 1 | 1 | 0% | 2,188 | 2,117 | -3% | 0 | 0 | — |
case-04 | pass→pass | 13,268 | 10,129 | -24% | 1 | 1 | 0% | 2,128 | 1,933 | -9% | 0 | 0 | — |
case-05 | fail→pass | 11,631 | 3,124 | -73% | 1 | 1 | 0% | 1,643 | 758 | -54% | 0 | 0 | — |
case-06 | fail→pass | 20,379 | 1,916 | -91% | 1 | 1 | 0% | 1,303 | 488 | -63% | 0 | 0 | — |
case-09 | pass→pass | 14,137 | 10,707 | -24% | 1 | 1 | 0% | 2,174 | 1,858 | -15% | 0 | 0 | — |
case-10 | pass→pass | 11,186 | 9,408 | -16% | 1 | 1 | 0% | 1,589 | 1,726 | +9% | 0 | 0 | — |
case-11 | fail→fail | 7,078 | 4,840 | -32% | 1 | 1 | 0% | 957 | 1,059 | +11% | 0 | 0 | — |
case-12 | fail→pass | 1,990 | 3,327 | +67% | 1 | 1 | 0% | 266 | 800 | +201% | 0 | 0 | — |
case-13 | pass→pass | 14,620 | 9,903 | -32% | 1 | 1 | 0% | 2,115 | 1,764 | -17% | 0 | 0 | — |
case-19 | fail→pass | 10,296 | 3,217 | -69% | 1 | 1 | 0% | 1,518 | 754 | -50% | 0 | 0 | — |
case-20 | fail→fail | 37,129 | 34,859 | -6% | 1 | 1 | 0% | 5,173 | 5,303 | +3% | 0 | 0 | — |
case-21 | fail→fail | 2,795 | 7,615 | +172% | 1 | 1 | 0% | 420 | 1,587 | +278% | 0 | 0 | — |
case-22 | fail→fail | 19,777 | 12,029 | -39% | 1 | 1 | 0% | 3,450 | 2,277 | -34% | 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, and 21 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 +41 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.