Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guidelines for updating citation counts for papers in the section files using the `update_citation_counts.py` tool. USE FOR: Updating citation counts for papers listed in the section files to keep information current. DO NOT USE FOR: 1) Adding new papers to the section files; 2) Classifying entries into sections.
.claude/skills/kimtth-update-cite-count/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -24% | 0% |
To keep the citation counts for papers in the section files up to date, follow these guidelines when using the update_citation_counts.py tool.
section/best_practices.md: RAG Research (Ranked by cite count >=100) and Agent Research (Ranked by cite count >=100).update_citation_counts.py directly to fetch the latest Semantic Scholar citation counts for papers already listed in those two sections. Use --dry-run first when reviewing a large update.section/best_practices.md to confirm the counts changed correctly and the ranked order is still accurate. If counts change enough to affect ordering, manually reorder the entries.powershell.venv\Scripts\python.exe code/update_citation_counts.py --dry-run .venv\Scripts\python.exe code/update_citation_counts.py
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | pass→pass | 9,078 | 2,125 | -77% | 1 | 1 | 0% | 1,318 | 576 | -56% | 0 | 0 | — |
case-08 | pass→pass | 8,406 | 1,480 | -82% | 1 | 1 | 0% | 1,302 | 438 | -66% | 0 | 0 | — |
case-01 | fail→fail | 7,352 | 5,885 | -20% | 1 | 1 | 0% | 1,294 | 457 | -65% | 0 | 0 | — |
case-02 | fail→fail | 11,808 | 6,611 | -44% | 1 | 1 | 0% | 2,027 | 672 | -67% | 0 | 0 | — |
case-03 | fail→fail | 5,386 | 5,837 | +8% | 1 | 1 | 0% | 740 | 644 | -13% | 0 | 0 | — |
case-04 | fail→pass | 10,149 | 4,262 | -58% | 1 | 1 | 0% | 1,776 | 934 | -47% | 0 | 0 | — |
case-05 | pass→pass | 8,081 | 2,127 | -74% | 1 | 1 | 0% | 1,159 | 605 | -48% | 0 | 0 | — |
case-06 | fail→pass | 14,580 | 1,927 | -87% | 1 | 1 | 0% | 2,079 | 515 | -75% | 0 | 0 | — |
case-07 | fail→pass | 2,753 | 1,713 | -38% | 1 | 1 | 0% | 408 | 517 | +27% | 0 | 0 | — |
case-09 | fail→pass | 3,112 | 1,244 | -60% | 1 | 1 | 0% | 418 | 413 | -1% | 0 | 0 | — |
case-10 | fail→fail | 10,107 | 2,148 | -79% | 1 | 1 | 0% | 1,522 | 485 | -68% | 0 | 0 | — |
case-11 | fail→pass | 5,874 | 2,978 | -49% | 1 | 1 | 0% | 946 | 723 | -24% | 0 | 0 | — |
case-12 | fail→pass | 7,366 | 1,721 | -77% | 1 | 1 | 0% | 1,053 | 511 | -51% | 0 | 0 | — |
case-13 | fail→pass | 9,612 | 3,475 | -64% | 1 | 1 | 0% | 1,492 | 747 | -50% | 0 | 0 | — |
case-18 | pass→pass | 8,694 | 2,922 | -66% | 1 | 1 | 0% | 1,399 | 738 | -47% | 0 | 0 | — |
case-14 | fail→pass | 7,082 | 1,586 | -78% | 1 | 1 | 0% | 1,023 | 466 | -54% | 0 | 0 | — |
case-15 | fail→pass | 16,628 | 1,561 | -91% | 1 | 1 | 0% | 2,539 | 465 | -82% | 0 | 0 | — |
case-16 | fail→pass | 9,892 | 2,157 | -78% | 1 | 1 | 0% | 1,469 | 567 | -61% | 0 | 0 | — |
case-17 | fail→pass | 11,847 | 2,692 | -77% | 1 | 1 | 0% | 1,865 | 676 | -64% | 0 | 0 | — |
case-20 | pass→pass | 6,562 | 2,535 | -61% | 1 | 1 | 0% | 993 | 739 | -26% | 0 | 0 | — |
case-21 | pass→pass | 4,951 | 2,726 | -45% | 1 | 1 | 0% | 746 | 634 | -15% | 0 | 0 | — |
case-22 | fail→pass | 10,576 | 7,557 | -29% | 1 | 1 | 0% | 1,752 | 1,517 | -13% | 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 19 counted toward the lift figure. The other 3 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 +55 percentage points is the difference between those two pass rates over the 19 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.