Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →SOP for auditing wikilink coverage — scans all edges and reports which source pages are missing [[dir/slug]] wikilinks to their targets.
.claude/skills/yogsoth-ai-wiki-edge-audit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 0% | 0% |
| case-20 | ✓→✗ | ▼ Worse | -74% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -48% | 0% |
| case-15 | ✓→✓ | ＝ Same ✓ | -39% | 0% |
Audit wikilink coverage across all edges. Identifies source pages that have edges in _edges.jsonl but lack the corresponding [[dir/slug]] wikilink in their body text.
vault_edge_audit
vault_edge_audit (zero parameters)total_edges, covered, missing_count, missing[]missing_count == 0: report full coverage, donemissing_count > 0: for each missing entry, open the source page and insert [[dir/slug]] (where dir/slug = target path minus .md) at a semantically relevant location in the body. If no suitable location, append a sentence containing the wikilink at end of body.vault_edge_audit to confirm coverage<HARD-GATE> Do NOT skip the fix step. Every missing wikilink must be resolved before reporting success. Re-audit must show missing_count == 0. </HARD-GATE>
Returns: { total_edges: number, covered: number, fixed: number }
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | pass→pass | 5,734 | 8,026 | +40% | 1 | 1 | 0% | 1,064 | 646 | -39% | 0 | 0 | — |
case-01 | fail→fail | 15,829 | 8,573 | -46% | 1 | 1 | 0% | 391 | 607 | +55% | 0 | 0 | — |
case-02 | fail→fail | 13,916 | 8,075 | -42% | 1 | 1 | 0% | 172 | 648 | +277% | 0 | 0 | — |
case-03 | fail→fail | 13,009 | 8,785 | -32% | 1 | 1 | 0% | 235 | 600 | +155% | 0 | 0 | — |
case-04 | fail→fail | 22,532 | 9,079 | -60% | 1 | 1 | 0% | 1,853 | 553 | -70% | 0 | 0 | — |
case-05 | pass→fail | 9,976 | 14,763 | +48% | 1 | 1 | 0% | 741 | 739 | -0% | 0 | 0 | — |
case-06 | fail→fail | 6,820 | 7,744 | +14% | 1 | 1 | 0% | 1,012 | 748 | -26% | 0 | 0 | — |
case-07 | fail→fail | 11,303 | 7,901 | -30% | 1 | 1 | 0% | 1,013 | 739 | -27% | 0 | 0 | — |
case-08 | fail→pass | 5,143 | 11,931 | +132% | 1 | 1 | 0% | 912 | 1,283 | +41% | 0 | 0 | — |
case-09 | pass→pass | 10,095 | 7,424 | -26% | 1 | 1 | 0% | 1,444 | 722 | -50% | 0 | 0 | — |
case-10 | pass→pass | 12,667 | 8,409 | -34% | 1 | 1 | 0% | 1,329 | 672 | -49% | 0 | 0 | — |
case-11 | fail→fail | 15,651 | 4,620 | -70% | 1 | 1 | 0% | 1,771 | 1,156 | -35% | 0 | 0 | — |
case-12 | pass→pass | 12,325 | 8,036 | -35% | 1 | 1 | 0% | 1,232 | 729 | -41% | 0 | 0 | — |
case-13 | pass→pass | 8,145 | 8,522 | +5% | 1 | 1 | 0% | 1,290 | 888 | -31% | 0 | 0 | — |
case-14 | fail→fail | 8,840 | 7,157 | -19% | 1 | 1 | 0% | 1,550 | 648 | -58% | 0 | 0 | — |
case-16 | fail→fail | 14,127 | 22,147 | +57% | 1 | 1 | 0% | 1,493 | 1,856 | +24% | 0 | 0 | — |
case-17 | pass→pass | 11,461 | 7,231 | -37% | 1 | 1 | 0% | 1,160 | 610 | -47% | 0 | 0 | — |
case-18 | pass→pass | 5,121 | 2,854 | -44% | 1 | 1 | 0% | 794 | 827 | +4% | 0 | 0 | — |
case-19 | fail→fail | 16,109 | 6,996 | -57% | 1 | 1 | 0% | 2,607 | 531 | -80% | 0 | 0 | — |
case-20 | pass→fail | 25,098 | 17,739 | -29% | 1 | 1 | 0% | 3,823 | 1,008 | -74% | 0 | 0 | — |
case-21 | pass→fail | 18,450 | 17,370 | -6% | 1 | 1 | 0% | 2,794 | 1,454 | -48% | 0 | 0 | — |
case-22 | fail→fail | 11,045 | 14,297 | +29% | 1 | 1 | 0% | 280 | 1,171 | +318% | 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 14 counted toward the lift figure. The other 8 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 -9 percentage points is the difference between those two pass rates over the 14 comparable cases. 4 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.