Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build and maintain a research repository that makes findings findable, reusable, and cumulative across the organization.
.claude/skills/owl-listener-research-repository/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 64% | 0% |
You are an expert in organizing research so it compounds in value rather than disappearing into shared drives.
You design and maintain the systems, tagging conventions, and rituals that keep research findable and used — so teams don't repeat studies, can build on prior work, and can make decisions backed by accumulated evidence.
Most research is conducted well and then effectively lost. Common failure modes:
Design the repository so insights are the primary entry point — not studies, not raw data.
Each insight should have:
The tagging system is the most critical design decision in a repository. Define tags before populating:
onboarding not Onboarding or onboardA repository is only as good as the habits around it:
Common tools used as research repositories: | Tool | Strengths | Weaknesses | |---|---|---| | Notion | Flexible structure, links, good search | Requires disciplined setup; search is approximate | | Airtable | Strong filtering, tagging, views | Less natural for narrative content | | Dovetail | Purpose-built for research; tagging + transcripts | Cost; another tool for teams to adopt | | Confluence | Integrated with Jira workflows | Poor search; hard to browse by insight | | EnjoyHQ | Purpose-built; good tagging | Cost; less common | The tool matters less than the structure and tagging conventions — a well-maintained Notion is more useful than a poorly-maintained Dovetail.
Test the repository's usefulness with these questions before considering it functional:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 20,010 | 26,813 | +34% | 1 | 1 | 0% | 3,143 | 5,469 | +74% | 0 | 0 | — |
case-02 | fail→fail | 15,312 | 18,973 | +24% | 1 | 1 | 0% | 2,822 | 4,314 | +53% | 0 | 0 | — |
case-03 | fail→pass | 20,565 | 22,530 | +10% | 1 | 1 | 0% | 3,640 | 5,258 | +44% | 0 | 0 | — |
case-04 | pass→pass | 12,415 | 12,280 | -1% | 1 | 1 | 0% | 2,143 | 3,195 | +49% | 0 | 0 | — |
case-05 | pass→pass | 10,731 | 10,763 | +0% | 1 | 1 | 0% | 1,820 | 2,888 | +59% | 0 | 0 | — |
case-06 | fail→pass | 12,573 | 10,684 | -15% | 1 | 1 | 0% | 2,135 | 2,974 | +39% | 0 | 0 | — |
case-07 | fail→pass | 11,558 | 9,842 | -15% | 1 | 1 | 0% | 1,862 | 2,720 | +46% | 0 | 0 | — |
case-08 | fail→pass | 12,748 | 13,694 | +7% | 1 | 1 | 0% | 2,070 | 3,387 | +64% | 0 | 0 | — |
case-09 | fail→pass | 13,488 | 14,717 | +9% | 1 | 1 | 0% | 2,319 | 3,338 | +44% | 0 | 0 | — |
case-10 | pass→pass | 13,097 | 12,907 | -1% | 1 | 1 | 0% | 2,185 | 3,120 | +43% | 0 | 0 | — |
case-11 | pass→pass | 13,689 | 13,729 | +0% | 1 | 1 | 0% | 2,066 | 3,276 | +59% | 0 | 0 | — |
case-12 | pass→pass | 12,794 | 8,655 | -32% | 1 | 1 | 0% | 1,982 | 2,407 | +21% | 0 | 0 | — |
case-13 | pass→pass | 15,370 | 14,337 | -7% | 1 | 1 | 0% | 2,475 | 3,512 | +42% | 0 | 0 | — |
case-14 | fail→pass | 12,655 | 6,917 | -45% | 1 | 1 | 0% | 2,283 | 2,338 | +2% | 0 | 0 | — |
case-15 | pass→pass | 12,565 | 10,423 | -17% | 1 | 1 | 0% | 2,158 | 2,821 | +31% | 0 | 0 | — |
case-16 | fail→pass | 9,126 | 10,561 | +16% | 1 | 1 | 0% | 1,555 | 2,770 | +78% | 0 | 0 | — |
case-17 | pass→pass | 12,323 | 14,845 | +20% | 1 | 1 | 0% | 2,016 | 3,453 | +71% | 0 | 0 | — |
case-18 | pass→pass | 16,524 | 16,313 | -1% | 1 | 1 | 0% | 2,592 | 3,547 | +37% | 0 | 0 | — |
case-19 | pass→pass | 4,820 | 6,022 | +25% | 1 | 1 | 0% | 777 | 2,150 | +177% | 0 | 0 | — |
case-20 | pass→pass | 15,393 | 15,773 | +2% | 1 | 1 | 0% | 2,875 | 3,798 | +32% | 0 | 0 | — |
case-21 | pass→pass | 14,325 | 14,311 | -0% | 1 | 1 | 0% | 2,950 | 4,072 | +38% | 0 | 0 | — |
case-22 | pass→pass | 17,931 | 14,961 | -17% | 1 | 1 | 0% | 2,941 | 3,386 | +15% | 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.
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.