Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →ESG (Environmental, Social, Governance) reporting and analytics. SASB, TCFD, GRI, and CDP frameworks. Carbon accounting, supply chain sustainability, diversity metrics, and regulatory disclosure support.
.claude/skills/mkurman-esg-reporting/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 76% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -15% | 0% |
| ----- | ---- | ------ | ------ | ------ | ---------- Scope 2 electricity | 48.0 | tCO2e | utility invoices | FY2025 | location-based factor | high Board independence | 4/5 | directors | board register | FY2025 | direct count | high Employee turnover | 12.4 | % | HRIS export | FY2025 | voluntary+involuntary | medium
## Writing rules
- Never present estimated numbers as measured numbers.
- Say which framework and year/version you are aligning to.
- Keep methodology notes near the metric.
- Separate commitments/goals from achieved results.
- If data coverage is partial, state boundary limitations explicitly.
## Common failure modes
- mixing entities or time periods
- using inconsistent emissions factors
- vague claims like “sustainable” without evidence
- reporting percentages with no denominator
- copying framework language without mapping to actual evidence| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→pass | 11,977 | 10,935 | -9% | 1 | 1 | 0% | 1,939 | 1,952 | +1% | 0 | 0 | — |
case-01 | fail→pass | 11,971 | 13,089 | +9% | 1 | 1 | 0% | 1,477 | 2,593 | +76% | 0 | 0 | — |
case-02 | fail→pass | 13,769 | 11,348 | -18% | 1 | 1 | 0% | 2,328 | 2,304 | -1% | 0 | 0 | — |
case-03 | fail→pass | 12,071 | 10,455 | -13% | 1 | 1 | 0% | 2,307 | 2,057 | -11% | 0 | 0 | — |
case-04 | pass→fail | 15,267 | 13,737 | -10% | 1 | 1 | 0% | 2,556 | 2,915 | +14% | 0 | 0 | — |
case-05 | pass→pass | 19,275 | 15,567 | -19% | 1 | 1 | 0% | 3,105 | 2,669 | -14% | 0 | 0 | — |
case-06 | pass→pass | 17,071 | 19,382 | +14% | 1 | 1 | 0% | 2,957 | 3,510 | +19% | 0 | 0 | — |
case-07 | pass→pass | 12,708 | 11,516 | -9% | 1 | 1 | 0% | 2,316 | 2,337 | +1% | 0 | 0 | — |
case-08 | fail→pass | 16,702 | 10,992 | -34% | 1 | 1 | 0% | 2,494 | 2,126 | -15% | 0 | 0 | — |
case-09 | pass→pass | 14,047 | 8,455 | -40% | 1 | 1 | 0% | 2,239 | 1,664 | -26% | 0 | 0 | — |
case-10 | pass→pass | 9,774 | 9,393 | -4% | 1 | 1 | 0% | 1,583 | 1,794 | +13% | 0 | 0 | — |
case-11 | pass→pass | 13,074 | 12,181 | -7% | 1 | 1 | 0% | 1,979 | 2,145 | +8% | 0 | 0 | — |
case-12 | pass→fail | 8,482 | 10,460 | +23% | 1 | 1 | 0% | 1,540 | 2,123 | +38% | 0 | 0 | — |
case-13 | pass→pass | 10,686 | 11,348 | +6% | 1 | 1 | 0% | 1,556 | 1,997 | +28% | 0 | 0 | — |
case-14 | fail→pass | 9,007 | 8,569 | -5% | 1 | 1 | 0% | 1,285 | 1,573 | +22% | 0 | 0 | — |
case-16 | pass→pass | 12,673 | 9,424 | -26% | 1 | 1 | 0% | 2,112 | 1,801 | -15% | 0 | 0 | — |
case-17 | pass→pass | 6,632 | 4,737 | -29% | 1 | 1 | 0% | 1,119 | 991 | -11% | 0 | 0 | — |
case-18 | pass→pass | 9,848 | 8,176 | -17% | 1 | 1 | 0% | 1,831 | 1,642 | -10% | 0 | 0 | — |
case-19 | pass→pass | 9,483 | 9,134 | -4% | 1 | 1 | 0% | 1,935 | 1,817 | -6% | 0 | 0 | — |
case-20 | pass→pass | 9,733 | 10,207 | +5% | 1 | 1 | 0% | 1,712 | 1,854 | +8% | 0 | 0 | — |
case-21 | pass→pass | 12,281 | 11,513 | -6% | 1 | 1 | 0% | 2,092 | 2,091 | -0% | 0 | 0 | — |
case-22 | pass→pass | 12,292 | 8,233 | -33% | 1 | 1 | 0% | 1,947 | 1,613 | -17% | 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 +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 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.