Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scans the knowledge base for entities that appear together in multiple contexts but have no explicit edge in the knowledge graph. These are implicit relationships that may deserve explicit documentation. The `--min` flag sets the co-occurrence threshold.
.claude/skills/miosa-osa-reflect/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -89% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -65% | 0% |
Scan for entity co-occurrences not captured as graph edges.
Scans entities that co-occur in contexts but have no edge between them. Suggests missing relationships. With Ollama: classifies relationship type (works_with, reports_to, related_to, depends_on). Without: reports co-occurrence count.
/reflect # Scan with default limit
/reflect --limit 20 # More results| Flag | Description | Default | |------|-------------|---------| | --limit | Max suggestions | 10 |
bashcd engine && mix optimal.reflect cd engine && mix optimal.reflect --limit 20
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 12,624 | 5,176 | -59% | 1 | 1 | 0% | 1,827 | 877 | -52% | 0 | 0 | — |
case-02 | fail→pass | 13,807 | 3,501 | -75% | 1 | 1 | 0% | 1,842 | 657 | -64% | 0 | 0 | — |
case-03 | fail→pass | 3,757 | 3,570 | -5% | 1 | 1 | 0% | 469 | 706 | +51% | 0 | 0 | — |
case-04 | fail→pass | 25,267 | 2,251 | -91% | 1 | 1 | 0% | 3,392 | 389 | -89% | 0 | 0 | — |
case-05 | fail→pass | 21,792 | 3,011 | -86% | 1 | 1 | 0% | 1,623 | 576 | -65% | 0 | 0 | — |
case-06 | pass→pass | 14,901 | 3,900 | -74% | 1 | 1 | 0% | 2,339 | 872 | -63% | 0 | 0 | — |
case-07 | fail→pass | 11,872 | 2,372 | -80% | 1 | 1 | 0% | 1,601 | 575 | -64% | 0 | 0 | — |
case-08 | pass→pass | 10,081 | 2,531 | -75% | 1 | 1 | 0% | 1,603 | 516 | -68% | 0 | 0 | — |
case-09 | pass→pass | 13,117 | 2,468 | -81% | 1 | 1 | 0% | 2,039 | 512 | -75% | 0 | 0 | — |
case-10 | pass→pass | 10,737 | 2,656 | -75% | 1 | 1 | 0% | 1,582 | 590 | -63% | 0 | 0 | — |
case-11 | pass→pass | 13,366 | 2,548 | -81% | 1 | 1 | 0% | 2,089 | 594 | -72% | 0 | 0 | — |
case-12 | fail→pass | 40,198 | 1,880 | -95% | 1 | 1 | 0% | 6,738 | 434 | -94% | 0 | 0 | — |
case-13 | fail→pass | 24,497 | 1,511 | -94% | 1 | 1 | 0% | 3,986 | 409 | -90% | 0 | 0 | — |
case-14 | fail→pass | 13,033 | 2,238 | -83% | 1 | 1 | 0% | 2,099 | 449 | -79% | 0 | 0 | — |
case-15 | fail→pass | 10,791 | 5,164 | -52% | 1 | 1 | 0% | 1,684 | 933 | -45% | 0 | 0 | — |
case-16 | pass→pass | 14,440 | 8,144 | -44% | 1 | 1 | 0% | 1,928 | 1,398 | -27% | 0 | 0 | — |
case-17 | fail→pass | 13,447 | 3,214 | -76% | 1 | 1 | 0% | 1,999 | 697 | -65% | 0 | 0 | — |
case-18 | fail→pass | 10,139 | 3,612 | -64% | 1 | 1 | 0% | 1,336 | 788 | -41% | 0 | 0 | — |
case-19 | pass→pass | 12,921 | 3,182 | -75% | 1 | 1 | 0% | 1,970 | 383 | -81% | 0 | 0 | — |
case-20 | fail→pass | 11,252 | 1,255 | -89% | 1 | 1 | 0% | 1,852 | 379 | -80% | 0 | 0 | — |
case-21 | pass→pass | 13,240 | 7,572 | -43% | 1 | 1 | 0% | 2,085 | 1,521 | -27% | 0 | 0 | — |
case-22 | pass→pass | 17,976 | 12,046 | -33% | 1 | 1 | 0% | 3,107 | 2,253 | -27% | 0 | 0 | — |
case-23 | pass→pass | 9,982 | 6,012 | -40% | 1 | 1 | 0% | 1,543 | 1,173 | -24% | 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. 23 cases were attempted. The headline lift of +57 percentage points is the difference between those two pass rates over the 23 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.