Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Map causal mechanisms between variables
.claude/skills/yogsoth-ai-mechanism-mapping/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 8% | 0% |
Construct the directed causal graph by creating mechanism edges between the variables identified in the variable-identification phase. Each edge must carry a plain-language description of the underlying mechanism and at least one evidence citation.
CC must treat each proposed edge as a hypothesis, not a fact. For every mechanism edge created, the description should answer three questions: what is the pathway, under what conditions does it hold, and what would falsify it. Edges without a plausible mechanism description are placeholders and must be revisited. The goal is a graph where every arrow is defensible, not merely plausible.
| Metric | S | M | L | |--------|---|---|---| | Mechanism edges created | 10 | 25 | 50 | | Mechanism descriptions | 5 | 15 | 30 | | Evidence citations | 8 | 20 | 40 |
| Metric | Target | Current | Status |
|-------------------------|--------|---------|--------|
| Mechanism edges created | S:10 / M:25 / L:50 | 0 | ⬜ |
| Mechanism descriptions | S:5 / M:15 / L:30 | 0 | ⬜ |
| Evidence citations | S:8 / M:20 / L:40 | 0 | ⬜ |<HARD-GATE> Cannot exit until 80% of budget met. Print state ledger before each iteration decision. </HARD-GATE>
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | counterfactual-reasoning | Tactic for reasoning about what would happen if variables were different — supports causal identification and intervention analysis. | | evidence-weighing | Tactic for assessing the strength and relevance of evidence for causal claims — distinguishes correlation from causation. | | feedback-loop-detection | Tactic for identifying circular causation — detect feedback loops, classify as reinforcing or balancing, document loop structure. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | mechanism-edge-creation | SOP for creating a causal mechanism edge — documents how one variable causes changes in another. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 36,906 | 32,144 | -13% | 1 | 1 | 0% | 2,213 | 4,321 | +95% | 0 | 0 | — |
case-06 | pass→fail | 34,634 | 60,466 | +75% | 1 | 1 | 0% | 3,454 | 6,403 | +85% | 0 | 0 | — |
case-07 | fail→pass | 62,540 | 51,451 | -18% | 1 | 1 | 0% | 8,258 | 8,884 | +8% | 0 | 0 | — |
case-01 | fail→pass | 51,813 | 64,002 | +24% | 1 | 1 | 0% | 8,288 | 8,914 | +8% | 0 | 0 | — |
case-02 | fail→fail | 44,720 | 56,390 | +26% | 1 | 1 | 0% | 8,269 | 8,895 | +8% | 0 | 0 | — |
case-03 | fail→pass | 62,050 | 65,115 | +5% | 1 | 1 | 0% | 8,276 | 8,902 | +8% | 0 | 0 | — |
case-04 | fail→fail | 14,098 | 49,169 | +249% | 1 | 1 | 0% | 1,630 | 8,869 | +444% | 0 | 0 | — |
case-08 | fail→pass | 58,533 | 54,621 | -7% | 1 | 1 | 0% | 6,880 | 8,886 | +29% | 0 | 0 | — |
case-09 | pass→pass | 49,064 | 66,163 | +35% | 1 | 1 | 0% | 4,100 | 7,255 | +77% | 0 | 0 | — |
case-10 | fail→fail | 70,206 | 69,431 | -1% | 1 | 1 | 0% | 7,988 | 8,877 | +11% | 0 | 0 | — |
case-11 | pass→fail | 50,330 | 50,312 | -0% | 1 | 1 | 0% | 8,247 | 8,873 | +8% | 0 | 0 | — |
case-12 | fail→pass | 48,392 | 60,272 | +25% | 1 | 1 | 0% | 8,246 | 8,872 | +8% | 0 | 0 | — |
case-13 | pass→pass | 36,194 | 50,177 | +39% | 1 | 1 | 0% | 5,565 | 8,868 | +59% | 0 | 0 | — |
case-14 | fail→fail | 46,696 | 47,454 | +2% | 1 | 1 | 0% | 8,254 | 8,880 | +8% | 0 | 0 | — |
case-15 | fail→pass | 27,536 | 50,336 | +83% | 1 | 1 | 0% | 4,263 | 8,866 | +108% | 0 | 0 | — |
case-16 | pass→pass | 50,628 | 46,552 | -8% | 1 | 1 | 0% | 7,251 | 8,869 | +22% | 0 | 0 | — |
case-17 | pass→pass | 35,129 | 51,524 | +47% | 1 | 1 | 0% | 6,246 | 8,867 | +42% | 0 | 0 | — |
case-18 | pass→pass | 43,136 | 60,760 | +41% | 1 | 1 | 0% | 6,512 | 8,868 | +36% | 0 | 0 | — |
case-19 | fail→pass | 35,708 | 60,938 | +71% | 1 | 1 | 0% | 5,379 | 8,871 | +65% | 0 | 0 | — |
case-20 | pass→pass | 42,063 | 37,592 | -11% | 1 | 1 | 0% | 8,244 | 7,626 | -7% | 0 | 0 | — |
case-21 | fail→pass | 42,174 | 53,967 | +28% | 1 | 1 | 0% | 7,210 | 8,607 | +19% | 0 | 0 | — |
case-22 | pass→pass | 39,767 | 46,309 | +16% | 1 | 1 | 0% | 8,238 | 8,864 | +8% | 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 +27 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.