Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Keep a team knowledge base alive — the gardener role and its weekly half-hour, the rot signals (stale pages, orphans, duplicates) and their fixes, the capture funnels that feed the garden, and the pruning that keeps search useful. Use when asked our wiki is a graveyard, who maintains the knowledge base, set up knowledge management that lasts, or people can't find anything anymore. Produces the gardener rotation, the weekly tending routine, the rot triage, and the capture funnels.
.claude/skills/mohitagw15856-knowledge-gardening/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 108% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 53% | 0% |
Knowledge bases don't die of bad writing — they die of no gardener: pages rot silently, duplicates sprout (because the original wasn't findable), orphans accumulate, and within two years search returns three contradictory versions of everything, teaching everyone to ask humans again. The fix isn't a heroic cleanup (that's the shared-drive-cleanup move, needed once) — it's gardening: a named rotating role, a weekly half-hour routine, rot signals with standard fixes, and capture funnels that turn the team's answered-questions and decisions into pages while they're fresh.
Ask for these if not provided:
Rotation calendar · the 30-minute scope · the visible weekly note format]
Triage new → fix flagged → prune one → merge one · with the platform's tools named]
| Signal | Standard fix | |---|---|
Answered-twice → pages · decision-log links · handoff/offboarding harvests — each with its trigger]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,449 | 21,295 | +9% | 1 | 1 | 0% | 3,127 | 4,628 | +48% | 0 | 0 | — |
case-02 | fail→pass | 19,102 | 19,019 | -0% | 1 | 1 | 0% | 3,221 | 4,548 | +41% | 0 | 0 | — |
case-03 | fail→fail | 20,000 | 21,424 | +7% | 1 | 1 | 0% | 3,450 | 4,833 | +40% | 0 | 0 | — |
case-04 | pass→pass | 17,083 | 18,468 | +8% | 1 | 1 | 0% | 2,895 | 4,319 | +49% | 0 | 0 | — |
case-05 | fail→fail | 15,409 | 14,541 | -6% | 1 | 1 | 0% | 2,447 | 3,773 | +54% | 0 | 0 | — |
case-06 | fail→fail | 19,216 | 12,683 | -34% | 1 | 1 | 0% | 3,226 | 3,507 | +9% | 0 | 0 | — |
case-07 | fail→pass | 14,942 | 12,823 | -14% | 1 | 1 | 0% | 2,459 | 3,327 | +35% | 0 | 0 | — |
case-08 | pass→pass | 11,782 | 8,978 | -24% | 1 | 1 | 0% | 1,844 | 2,611 | +42% | 0 | 0 | — |
case-09 | fail→pass | 8,846 | 11,282 | +28% | 1 | 1 | 0% | 1,450 | 3,021 | +108% | 0 | 0 | — |
case-10 | fail→pass | 8,366 | 6,229 | -26% | 1 | 1 | 0% | 1,435 | 2,198 | +53% | 0 | 0 | — |
case-11 | pass→pass | 13,519 | 8,680 | -36% | 1 | 1 | 0% | 2,236 | 2,496 | +12% | 0 | 0 | — |
case-12 | pass→pass | 5,808 | 3,228 | -44% | 1 | 1 | 0% | 927 | 1,746 | +88% | 0 | 0 | — |
case-13 | fail→pass | 10,157 | 6,328 | -38% | 1 | 1 | 0% | 1,715 | 2,243 | +31% | 0 | 0 | — |
case-14 | pass→pass | 10,645 | 7,828 | -26% | 1 | 1 | 0% | 1,681 | 2,442 | +45% | 0 | 0 | — |
case-15 | pass→pass | 16,469 | 15,441 | -6% | 1 | 1 | 0% | 2,722 | 3,616 | +33% | 0 | 0 | — |
case-16 | pass→pass | 15,276 | 13,247 | -13% | 1 | 1 | 0% | 2,320 | 3,209 | +38% | 0 | 0 | — |
case-17 | fail→fail | 12,545 | 11,710 | -7% | 1 | 1 | 0% | 2,083 | 3,161 | +52% | 0 | 0 | — |
case-18 | pass→pass | 15,141 | 13,853 | -9% | 1 | 1 | 0% | 2,433 | 3,442 | +41% | 0 | 0 | — |
case-19 | pass→pass | 11,826 | 6,557 | -45% | 1 | 1 | 0% | 2,027 | 2,193 | +8% | 0 | 0 | — |
case-20 | fail→pass | 13,901 | 7,918 | -43% | 1 | 1 | 0% | 2,380 | 2,570 | +8% | 0 | 0 | — |
case-21 | pass→pass | 12,703 | 7,276 | -43% | 1 | 1 | 0% | 2,041 | 2,315 | +13% | 0 | 0 | — |
case-22 | fail→fail | 10,999 | 5,435 | -51% | 1 | 1 | 0% | 1,794 | 2,303 | +28% | 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 +32 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 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.