Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run a quarterly loan covenant compliance review: covenant table with required vs actual vs headroom, trend and trajectory-to-breach analysis, waiver and amendment options with pricing implications, early-warning indicators, and a watch-list recommendation. Use when asked to review covenant compliance, check covenant headroom, assess a potential covenant breach, or prepare a quarterly borrower monitoring review. Produces a structured covenant review with headroom table, trajectory analysis, and r
.claude/skills/mohitagw15856-loan-covenant-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 73% | 0% |
Covenants are the bank's smoke detectors — but only if someone reads them as trends, not snapshots. This skill runs the quarterly discipline: where is each covenant now, which direction is it moving, when does the trend cross the line, and what do we do before it does.
Ask for what's missing; with partial data, compute what's computable and mark the rest [awaiting financials]:
1. Covenant table. For each covenant: required level, actual, headroom % = (actual − required) ÷ required (sign-adjusted so positive = compliant for both maximum-leverage and minimum-coverage covenants). Recompute actuals from the definitions in the agreement — borrower certificates use borrower-friendly add-backs; note any definitional divergence.
2. Headroom bands:
| Headroom | Status | Response | |---|---|---| | >20% | Comfortable | Routine monitoring | | 10–20% | Monitor | Note trend; quarterly attention | | <10% | Early warning | Trajectory analysis, proactive borrower dialogue | | Breached | Breach | Reservation of rights, waiver/amendment track, risk-grade review |
3. Trajectory-to-breach. For anything under 20% or trending down 2+ quarters: extend the trend 2–4 quarters and state the projected breach quarter with the assumption ("on the last 3 quarters' EBITDA slope, leverage crosses 3.5x in Q2"). Label it a trend projection, not a forecast. Seasonality: compare year-on-year quarters before calling a trend.
4. Options if breach is likely or occurred. Frame the menu with typical pricing logic (calibrate to institution practice): one-off waiver (waiver fee, often bps on commitment); reset/amendment (amendment fee + margin step-up, often with tightened baskets, added reporting, or a sweep); equity cure if documented; standstill/reservation of rights while options are assessed. Note: accepting payment or staying silent after a known breach can prejudice rights — flag "reserve rights promptly" whenever a breach exists.
5. Early-warning indicators (qualitative). Check: late or requalified financials, auditor change or going-concern language, CFO/management turnover, maxed revolver utilisation creep, stretched payables, delayed compliance certificates, adverse sector news. Two or more present → recommend watch-list consideration even with numeric compliance.
1. Summary verdict — compliant / early warning / breach, one paragraph. 2. Covenant table — covenant | definition source | required | actual | headroom % | 4-quarter trend | status. 3. Trajectory analysis — projected breach quarter and assumptions, per tightening covenant. 4. Early-warning indicators — checklist with evidence. 5. Options & pricing implications — if relevant: waiver / amend / cure / reserve rights, with trade-offs. 6. Recommendation — grade/watch-list action, borrower conversation points, next review date.
End with: "This review is analytical support, not a credit or enforcement decision. Waivers, grading, and rights reservations follow your institution's credit policy and applicable regulation."
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→fail | 28,005 | 7,755 | -72% | 1 | 1 | 0% | 5,286 | 2,500 | -53% | 0 | 0 | — |
case-08 | fail→pass | 14,443 | 16,479 | +14% | 1 | 1 | 0% | 2,419 | 3,837 | +59% | 0 | 0 | — |
case-02 | fail→pass | 30,148 | 32,587 | +8% | 1 | 1 | 0% | 6,258 | 7,336 | +17% | 0 | 0 | — |
case-03 | fail→pass | 31,632 | 24,993 | -21% | 1 | 1 | 0% | 6,249 | 5,522 | -12% | 0 | 0 | — |
case-04 | fail→pass | 8,101 | 8,116 | +0% | 1 | 1 | 0% | 1,541 | 2,600 | +69% | 0 | 0 | — |
case-05 | pass→pass | 9,611 | 13,460 | +40% | 1 | 1 | 0% | 1,891 | 3,504 | +85% | 0 | 0 | — |
case-06 | fail→pass | 9,415 | 13,331 | +42% | 1 | 1 | 0% | 1,951 | 3,373 | +73% | 0 | 0 | — |
case-07 | pass→pass | 13,867 | 19,697 | +42% | 1 | 1 | 0% | 2,351 | 4,671 | +99% | 0 | 0 | — |
case-09 | pass→pass | 10,209 | 15,382 | +51% | 1 | 1 | 0% | 1,600 | 3,624 | +126% | 0 | 0 | — |
case-10 | fail→pass | 12,261 | 13,199 | +8% | 1 | 1 | 0% | 2,258 | 3,647 | +62% | 0 | 0 | — |
case-11 | pass→pass | 15,460 | 13,925 | -10% | 1 | 1 | 0% | 1,500 | 3,288 | +119% | 0 | 0 | — |
case-12 | fail→fail | 24,858 | 12,999 | -48% | 1 | 1 | 0% | 2,257 | 3,130 | +39% | 0 | 0 | — |
case-13 | fail→pass | 10,088 | 10,327 | +2% | 1 | 1 | 0% | 1,366 | 2,085 | +53% | 0 | 0 | — |
case-14 | fail→pass | 20,396 | 12,295 | -40% | 1 | 1 | 0% | 1,249 | 1,749 | +40% | 0 | 0 | — |
case-15 | pass→pass | 18,574 | 16,703 | -10% | 1 | 1 | 0% | 2,160 | 3,696 | +71% | 0 | 0 | — |
case-16 | pass→pass | 8,755 | 17,384 | +99% | 1 | 1 | 0% | 1,563 | 2,035 | +30% | 0 | 0 | — |
case-17 | pass→pass | 9,443 | 8,388 | -11% | 1 | 1 | 0% | 1,718 | 2,380 | +39% | 0 | 0 | — |
case-18 | pass→pass | 9,705 | 6,787 | -30% | 1 | 1 | 0% | 2,102 | 2,433 | +16% | 0 | 0 | — |
case-19 | pass→pass | 10,115 | 7,927 | -22% | 1 | 1 | 0% | 1,754 | 2,484 | +42% | 0 | 0 | — |
case-20 | pass→pass | 32,846 | 20,675 | -37% | 1 | 1 | 0% | 6,192 | 4,782 | -23% | 0 | 0 | — |
case-21 | pass→pass | 14,966 | 19,627 | +31% | 1 | 1 | 0% | 2,849 | 4,664 | +64% | 0 | 0 | — |
case-22 | pass→pass | 13,397 | 11,819 | -12% | 1 | 1 | 0% | 2,513 | 3,361 | +34% | 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 +32 percentage points is the difference between those two pass rates over the 22 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.