Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search the OD Library (the global asset registry) and apply matching assets into the current project mid-task. Use when the user asks to reuse an image they captured/uploaded earlier, "pull a logo/screenshot from my library", or to find and drop a stored asset into the page being built.
.claude/skills/nexu-io-library-curator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 123% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 614% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 120% | 0% |
Reuse assets that already live in the user's OD Library — images captured with the OD Clipper, manual uploads, agent-generated media, and design-system material — without asking the user to re-upload them.
"my logo", "that hero image from earlier").
library over freshly generated media.
Both endpoints authenticate with the run's tool token (OD_TOOL_TOKEN, injected by the daemon) and operate on the project the run belongs to.
POST /api/tools/library/search
json{ "query": "blue hero background", "kind": "image", "limit": 20 }
Returns { "results": [{ "asset": { "id": "...", "kind": "image", "sourceTitle": "...", "width": 1600, "height": 900, "sources": [...] }, "score": 0 }], "semantic": false }.
semantic: false means keyword/metadata matching (no embedding model configured). Filter and rank the results yourself from the asset metadata.
POST /api/tools/library/apply
json{ "assetId": "<id from search>", "dir": "assets" }
Copies the asset into the project (default subdir library/, or the dir you pass) and returns { "relPath": "assets/<hash>.png" }. Reference that relPath from the HTML/CSS you write (e.g. <img src="assets/ab12cd34ef.png">).
If search returns nothing, fall back to media generation rather than guessing a path — never invent a relPath that apply did not return.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 25,731 | 38,292 | +49% | 1 | 1 | 0% | 3,963 | 8,714 | +120% | 0 | 0 | — |
case-01 | fail→fail | 11,400 | 28,458 | +150% | 1 | 1 | 0% | 1,645 | 849 | -48% | 0 | 0 | — |
case-02 | fail→fail | 11,752 | 7,897 | -33% | 1 | 1 | 0% | 1,145 | 808 | -29% | 0 | 0 | — |
case-03 | fail→fail | 5,466 | 8,132 | +49% | 1 | 1 | 0% | 907 | 997 | +10% | 0 | 0 | — |
case-04 | pass→pass | 15,706 | 38,196 | +143% | 1 | 1 | 0% | 2,781 | 6,861 | +147% | 0 | 0 | — |
case-05 | fail→fail | 5,447 | 9,458 | +74% | 1 | 1 | 0% | 909 | 934 | +3% | 0 | 0 | — |
case-07 | fail→fail | 6,517 | 15,898 | +144% | 1 | 1 | 0% | 272 | 1,371 | +404% | 0 | 0 | — |
case-08 | fail→fail | 10,318 | 15,084 | +46% | 1 | 1 | 0% | 491 | 1,140 | +132% | 0 | 0 | — |
case-09 | fail→pass | 9,686 | 11,160 | +15% | 1 | 1 | 0% | 1,985 | 2,714 | +37% | 0 | 0 | — |
case-10 | fail→fail | 10,020 | 13,951 | +39% | 1 | 1 | 0% | 378 | 1,348 | +257% | 0 | 0 | — |
case-11 | pass→pass | 7,828 | 8,452 | +8% | 1 | 1 | 0% | 995 | 2,071 | +108% | 0 | 0 | — |
case-12 | fail→fail | 6,706 | 12,389 | +85% | 1 | 1 | 0% | 172 | 839 | +388% | 0 | 0 | — |
case-13 | fail→pass | 18,512 | 3,684 | -80% | 1 | 1 | 0% | 477 | 1,066 | +123% | 0 | 0 | — |
case-14 | fail→pass | 11,537 | 4,102 | -64% | 1 | 1 | 0% | 1,503 | 1,032 | -31% | 0 | 0 | — |
case-15 | fail→fail | 9,344 | 27,326 | +192% | 1 | 1 | 0% | 1,542 | 1,835 | +19% | 0 | 0 | — |
case-16 | pass→pass | 3,077 | 3,715 | +21% | 1 | 1 | 0% | 531 | 1,135 | +114% | 0 | 0 | — |
case-17 | fail→pass | 1,551 | 5,677 | +266% | 1 | 1 | 0% | 205 | 1,464 | +614% | 0 | 0 | — |
case-18 | fail→fail | 7,566 | 9,689 | +28% | 1 | 1 | 0% | 338 | 991 | +193% | 0 | 0 | — |
case-19 | fail→fail | 4,959 | 7,852 | +58% | 1 | 1 | 0% | 184 | 971 | +428% | 0 | 0 | — |
case-20 | fail→fail | 9,836 | 10,211 | +4% | 1 | 1 | 0% | 1,451 | 1,024 | -29% | 0 | 0 | — |
case-21 | fail→fail | 9,047 | 24,399 | +170% | 1 | 1 | 0% | 791 | 783 | -1% | 0 | 0 | — |
case-22 | pass→pass | 6,871 | 2,740 | -60% | 1 | 1 | 0% | 1,110 | 954 | -14% | 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 9 counted toward the lift figure. The other 13 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 +18 percentage points is the difference between those two pass rates over the 9 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.