Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use this skill when Zotero is the literature source of truth and the project KB should receive source notes under Sources/Papers plus project-linked synthesis in Knowledge and Writing.
.claude/skills/galaxy-dawn-zotero-obsidian-bridge/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 156% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -34% | 0% |
Use this skill when papers live in Zotero and the project KB should receive project-local notes.
Default flow:
textZotero -> Sources/Papers -> Knowledge -> Writing -> Maps/literature.canvas
Rules:
Sources/Papers/Knowledge/Writing/Maps/literature.canvas is the default derived graph artifact_system/registry.md, 02-Index.md, and today's Daily/ after substantial ingestionCore papers and papers with reusable claims should include an evidence record. Abstract-only or placeholder notes may stay in To-Read until there is enough evidence to extract a reusable claim.
md## Evidence Record Evidence ID: Source: Source type: full paper | preprint | dataset | experiment artifact | project note | abstract-only | webpage placeholder Supports: Contradicts: Method / dataset / metric: Limitation: Project relevance: Claim strength: speculative | observed | supported | strong
Use ../research-ideation/references/research-contract.md as the shared contract for Evidence Records and claim strength definitions.
Before synthesis lands in Knowledge/ or Writing/, apply the shared Claim Promotion Gate:
When synthesis lands in Knowledge/, link the supporting paper notes explicitly. Do not promote literature synthesis that cannot point back to source notes.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 8,578 | 4,703 | -45% | 1 | 1 | 0% | 1,288 | 1,226 | -5% | 0 | 0 | — |
case-02 | fail→pass | 5,678 | 10,142 | +79% | 1 | 1 | 0% | 847 | 2,170 | +156% | 0 | 0 | — |
case-03 | pass→pass | 10,836 | 3,752 | -65% | 1 | 1 | 0% | 1,805 | 1,032 | -43% | 0 | 0 | — |
case-04 | fail→pass | 11,816 | 49,513 | +319% | 1 | 1 | 0% | 2,214 | 1,206 | -46% | 0 | 0 | — |
case-09 | fail→pass | 14,988 | 2,873 | -81% | 1 | 1 | 0% | 2,439 | 971 | -60% | 0 | 0 | — |
case-05 | fail→pass | 8,638 | 3,666 | -58% | 1 | 1 | 0% | 1,597 | 1,048 | -34% | 0 | 0 | — |
case-06 | fail→pass | 5,222 | 4,564 | -13% | 1 | 1 | 0% | 878 | 658 | -25% | 0 | 0 | — |
case-07 | fail→pass | 12,288 | 2,903 | -76% | 1 | 1 | 0% | 2,054 | 948 | -54% | 0 | 0 | — |
case-08 | pass→pass | 10,705 | 5,716 | -47% | 1 | 1 | 0% | 1,662 | 1,408 | -15% | 0 | 0 | — |
case-10 | pass→pass | 12,308 | 21,045 | +71% | 1 | 1 | 0% | 1,939 | 745 | -62% | 0 | 0 | — |
case-11 | pass→pass | 10,882 | 4,341 | -60% | 1 | 1 | 0% | 1,671 | 1,257 | -25% | 0 | 0 | — |
case-12 | fail→pass | 7,960 | 1,037 | -87% | 1 | 1 | 0% | 1,311 | 578 | -56% | 0 | 0 | — |
case-13 | fail→pass | 6,298 | 2,829 | -55% | 1 | 1 | 0% | 964 | 899 | -7% | 0 | 0 | — |
case-14 | pass→pass | 14,659 | 8,153 | -44% | 1 | 1 | 0% | 2,271 | 1,823 | -20% | 0 | 0 | — |
case-15 | pass→pass | 9,073 | 4,836 | -47% | 1 | 1 | 0% | 1,380 | 1,250 | -9% | 0 | 0 | — |
case-16 | pass→pass | 14,316 | 2,263 | -84% | 1 | 1 | 0% | 2,342 | 770 | -67% | 0 | 0 | — |
case-17 | fail→pass | 6,984 | 1,933 | -72% | 1 | 1 | 0% | 1,067 | 704 | -34% | 0 | 0 | — |
case-18 | pass→pass | 8,668 | 3,437 | -60% | 1 | 1 | 0% | 1,340 | 967 | -28% | 0 | 0 | — |
case-19 | fail→pass | 10,498 | 2,249 | -79% | 1 | 1 | 0% | 1,284 | 769 | -40% | 0 | 0 | — |
case-20 | pass→pass | 22,227 | 2,166 | -90% | 1 | 1 | 0% | 1,386 | 769 | -45% | 0 | 0 | — |
case-21 | pass→pass | 70,691 | 4,158 | -94% | 1 | 1 | 0% | 1,418 | 1,194 | -16% | 0 | 0 | — |
case-22 | pass→pass | 8,428 | 10,302 | +22% | 1 | 1 | 0% | 1,319 | 1,951 | +48% | 0 | 0 | — |
case-23 | fail→fail | 7,830 | 6,628 | -15% | 1 | 1 | 0% | 1,331 | 1,492 | +12% | 0 | 0 | — |
case-24 | fail→fail | 16,570 | 14,903 | -10% | 1 | 1 | 0% | 3,739 | 3,534 | -5% | 0 | 0 | — |
case-25 | fail→fail | 9,728 | 7,489 | -23% | 1 | 1 | 0% | 1,898 | 1,885 | -1% | 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. 25 cases were attempted. The headline lift of +44 percentage points is the difference between those two pass rates over the 25 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.