Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Set a filename convention that sorts, searches, and survives — date-first ISO format, the descriptor grammar, version suffixes that end the FINAL-final2 era, and the rollout that gets a team actually using it. Use when asked set up file naming rules, our filenames are chaos, what should we call our files, or fix the v2-final-FINAL problem. Produces the convention with its grammar, examples for the team's real file types, the version rule, and the one-line cheat sheet.
.claude/skills/mohitagw15856-filename-convention/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 16% | 0% |
Report_final_v2_FINAL_johns-edits(2).docx is a filesystem crying for grammar. A working convention is small — date first in ISO form so files sort themselves, a few descriptor slots in fixed order, a version rule that bans "final" — and it's enforced by usefulness: names that sort chronologically and surface in search within seconds pay their own adoption. The convention fits on one line; everything longer goes unused.
YYYY-MM-DD_project_doctype_descriptor_v2) tuned to the team's real filesv1, v2, v3 + -draft/-approved states; "final" banned by decree with the reasonAsk for these if not provided:
acme_2026-07-19_... when browsing-by-client dominates); the retrieval pattern picks the lead slot2026-07-19 sorts correctly forever, is unambiguous across cultures, and makes every folder self-chronologizing. The one legitimate exception: entity-first when browsing-by-entity is the dominant retrieval — pick per the team's seek pattern, once.date_project_doctype_descriptor_version — underscores between slots, hyphens within them (acme-renewal). Fixed order means the eye parses filenames like a table; mixed order means reading every name in full.v1, v2), state tags where useful (v3-approved), and "final" is banned because it's a prophecy, not a state — v4-final-final2 is the ban's justification, printed on the cheat sheet. Platform-versioned files (cloud docs) drop the suffix entirely; duplicating the platform's version history in names is noise.board-pack, signed-copy, eu-pricing) — jargon the team shares is fine; codes only the author knows are not.YYYY-MM-DD_project_doctype_descriptor_vN — underscores between slots, hyphens inside, ISO dates, integers for versions, "final" is banned.
| Now | Under the convention | |---|---| The team's real dozen, renamed]
Cloud docs: title = name, no version suffix · exports/receipts: date+source+period · the entity-first variant if chosen, with its why]
Clients/Acme/acme-acme-contract wastes a slot| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,146 | 11,480 | -19% | 1 | 1 | 0% | 3,102 | 3,594 | +16% | 0 | 0 | — |
case-02 | fail→pass | 12,613 | 16,305 | +29% | 1 | 1 | 0% | 2,809 | 4,821 | +72% | 0 | 0 | — |
case-03 | fail→pass | 11,504 | 11,139 | -3% | 1 | 1 | 0% | 2,537 | 3,477 | +37% | 0 | 0 | — |
case-04 | pass→pass | 6,510 | 8,944 | +37% | 1 | 1 | 0% | 1,335 | 2,865 | +115% | 0 | 0 | — |
case-05 | pass→pass | 13,825 | 8,582 | -38% | 1 | 1 | 0% | 2,875 | 2,933 | +2% | 0 | 0 | — |
case-06 | fail→pass | 8,067 | 9,096 | +13% | 1 | 1 | 0% | 1,642 | 2,849 | +74% | 0 | 0 | — |
case-07 | fail→fail | 10,549 | 11,616 | +10% | 1 | 1 | 0% | 2,123 | 3,294 | +55% | 0 | 0 | — |
case-08 | fail→pass | 15,029 | 11,078 | -26% | 1 | 1 | 0% | 2,495 | 2,893 | +16% | 0 | 0 | — |
case-09 | pass→pass | 12,484 | 10,786 | -14% | 1 | 1 | 0% | 2,334 | 2,981 | +28% | 0 | 0 | — |
case-10 | pass→pass | 7,587 | 5,297 | -30% | 1 | 1 | 0% | 1,490 | 2,010 | +35% | 0 | 0 | — |
case-11 | pass→pass | 14,292 | 10,626 | -26% | 1 | 1 | 0% | 2,572 | 2,955 | +15% | 0 | 0 | — |
case-12 | pass→pass | 14,818 | 14,993 | +1% | 1 | 1 | 0% | 2,469 | 3,575 | +45% | 0 | 0 | — |
case-13 | pass→pass | 12,695 | 6,963 | -45% | 1 | 1 | 0% | 2,241 | 2,406 | +7% | 0 | 0 | — |
case-14 | pass→pass | 5,689 | 5,162 | -9% | 1 | 1 | 0% | 996 | 2,119 | +113% | 0 | 0 | — |
case-15 | fail→pass | 10,664 | 5,925 | -44% | 1 | 1 | 0% | 2,002 | 2,164 | +8% | 0 | 0 | — |
case-16 | pass→pass | 9,739 | 12,003 | +23% | 1 | 1 | 0% | 1,788 | 3,356 | +88% | 0 | 0 | — |
case-17 | pass→pass | 10,011 | 12,317 | +23% | 1 | 1 | 0% | 1,862 | 3,318 | +78% | 0 | 0 | — |
case-18 | fail→pass | 14,096 | 9,974 | -29% | 1 | 1 | 0% | 2,262 | 2,804 | +24% | 0 | 0 | — |
case-19 | fail→pass | 9,260 | 7,853 | -15% | 1 | 1 | 0% | 1,902 | 2,681 | +41% | 0 | 0 | — |
case-20 | pass→pass | 2,890 | 3,779 | +31% | 1 | 1 | 0% | 463 | 1,703 | +268% | 0 | 0 | — |
case-21 | pass→pass | 4,726 | 3,288 | -30% | 1 | 1 | 0% | 812 | 1,621 | +100% | 0 | 0 | — |
case-22 | pass→pass | 7,471 | 6,745 | -10% | 1 | 1 | 0% | 1,326 | 2,316 | +75% | 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 +36 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.