Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate master's thesis review reports (硕士论文评阅意见) calibrated to a given score. Outputs academic evaluation and shortcomings/suggestions in Chinese. Trigger when user says "master thesis review" / "硕士论文评阅" / "论文评阅" / "评阅意见" / "thesis review".
.claude/skills/brycewang-stanford-master-thesis-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 70% | 0% |
| case-09 | ✓→✗ | ▼ Worse | 9% | 0% |
| case-17 | ✓→✓ | = Same ✓ | 32% | 0% |
| case-18 | ✓→✓ | = Same ✓ | 37% | 0% |
| case-19 | ✓→✓ | = Same ✓ | 96% | 0% |
This skill generates a master's thesis review report (硕士学位论文评阅意见) for a given thesis. The user provides the thesis file and a score (out of 100). The skill reads the thesis, then generates two sections calibrated to the score level.
Input:
*.pdf, *.tex, or *.docx)Output: {original-filename}-评阅意见.doc saved in the project folder
$ARGUMENTS or ask: "请提供论文文件夹路径").*.pdf, *.tex, *.docx).Use one AskUserQuestion call with one question:
Question — 评分 (Score):
Record the exact score. If the user selects a range, use the midpoint (e.g., 90–100 → 95).
Generate two sections based on the thesis content and calibrated to the score.
The score determines the tone, praise-to-criticism ratio, and severity of issues raised:
要求:
要求:
-评阅意见.doc. Example: 张三_硕士论文.pdf → 张三_硕士论文-评阅意见.doc.对学位论文的学术评语
[240字学术评语段落]
论文的不足之处和建议
[300字不足与建议段落]
建议成绩:XX 分{filename}-评阅意见.doc。"| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→fail | 16,878 | 10,597 | -37% | 1 | 1 | 0% | 2,401 | 2,562 | +7% | 0 | 0 | — |
case-01 | fail→fail | 5,307 | 6,901 | +30% | 1 | 1 | 0% | 440 | 2,209 | +402% | 0 | 0 | — |
case-02 | fail→fail | 18,650 | 8,556 | -54% | 1 | 1 | 0% | 3,294 | 2,306 | -30% | 0 | 0 | — |
case-03 | fail→fail | 19,276 | 19,995 | +4% | 1 | 1 | 0% | 3,451 | 2,088 | -39% | 0 | 0 | — |
case-05 | fail→fail | 6,958 | 8,711 | +25% | 1 | 1 | 0% | 1,282 | 3,228 | +152% | 0 | 0 | — |
case-06 | fail→fail | 12,091 | 31,326 | +159% | 1 | 1 | 0% | 2,033 | 7,831 | +285% | 0 | 0 | — |
case-07 | fail→fail | 12,123 | 34,050 | +181% | 1 | 1 | 0% | 2,878 | 7,833 | +172% | 0 | 0 | — |
case-08 | fail→fail | 10,122 | 8,549 | -16% | 1 | 1 | 0% | 1,686 | 2,220 | +32% | 0 | 0 | — |
case-09 | pass→fail | 14,591 | 10,576 | -28% | 1 | 1 | 0% | 2,309 | 2,506 | +9% | 0 | 0 | — |
case-10 | fail→fail | 10,613 | 8,172 | -23% | 1 | 1 | 0% | 2,149 | 2,240 | +4% | 0 | 0 | — |
case-11 | fail→fail | 6,736 | 12,826 | +90% | 1 | 1 | 0% | 1,118 | 3,705 | +231% | 0 | 0 | — |
case-12 | fail→fail | 20,939 | 8,014 | -62% | 1 | 1 | 0% | 3,181 | 2,230 | -30% | 0 | 0 | — |
case-13 | fail→fail | 9,528 | 8,143 | -15% | 1 | 1 | 0% | 1,485 | 2,187 | +47% | 0 | 0 | — |
case-14 | fail→pass | 10,129 | 9,629 | -5% | 1 | 1 | 0% | 1,683 | 2,857 | +70% | 0 | 0 | — |
case-15 | fail→fail | 7,649 | 10,429 | +36% | 1 | 1 | 0% | 1,156 | 2,427 | +110% | 0 | 0 | — |
case-16 | fail→fail | 8,884 | 7,738 | -13% | 1 | 1 | 0% | 1,473 | 2,252 | +53% | 0 | 0 | — |
case-17 | pass→pass | 20,510 | 17,628 | -14% | 1 | 1 | 0% | 3,360 | 4,420 | +32% | 0 | 0 | — |
case-18 | pass→pass | 19,645 | 16,526 | -16% | 1 | 1 | 0% | 3,220 | 4,406 | +37% | 0 | 0 | — |
case-19 | pass→pass | 19,600 | 26,784 | +37% | 1 | 1 | 0% | 3,135 | 6,130 | +96% | 0 | 0 | — |
case-20 | fail→fail | 8,995 | 6,463 | -28% | 1 | 1 | 0% | 1,410 | 2,070 | +47% | 0 | 0 | — |
case-21 | fail→fail | 9,432 | 3,674 | -61% | 1 | 1 | 0% | 1,546 | 2,332 | +51% | 0 | 0 | — |
case-22 | fail→fail | 10,745 | 7,501 | -30% | 1 | 1 | 0% | 1,663 | 2,077 | +25% | 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 9 counted toward the lift figure. The other 13 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 0 percentage points is the difference between those two pass rates over the 9 comparable cases. 5 cases got worse with the skill loaded, and they are 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.