Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tracks free trials in memory and warns the user on Telegram before they get charged. The user adds a trial with "trial: [service], [days] days"; the agent stores the end date and runs a daily check, sending a 5-day warning, a 2-day urgent alert, and an expiry flag — so a forgotten trial never turns into a surprise charge.
.claude/skills/nearai-free-trial-guardian/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 200% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 207% | 0% |
| case-18 | ✓→✗ | ▼ Worse | -9% | 0% |
You guard the user's free trials so they never get charged for one they forgot to cancel.
time tool. Never assume today's date or hardcode it.HEARTBEAT_OK and stop.When the user says trial: [service], [days] days (e.g. trial: Netflix, 14 days):
trials/active.md with memory_read.time tool), end date (today + days), status ACTIVE.memory_write.Create a routine that runs every day at 9:00 AM:
trials/active.md.time tool.HEARTBEAT_OK and stop.Alert format: ⚠️ Free Trial Alert 🔴 EXPIRES IN 2 DAYS:
🟡 EXPIRES IN 5 DAYS:
⛔ JUST EXPIRED:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,512 | 4,258 | -23% | 1 | 1 | 0% | 919 | 620 | -33% | 0 | 0 | — |
case-02 | fail→fail | 4,675 | 4,848 | +4% | 1 | 1 | 0% | 707 | 619 | -12% | 0 | 0 | — |
case-03 | fail→fail | 3,317 | 3,778 | +14% | 1 | 1 | 0% | 495 | 814 | +64% | 0 | 0 | — |
case-04 | pass→pass | 4,361 | 2,888 | -34% | 1 | 1 | 0% | 723 | 1,043 | +44% | 0 | 0 | — |
case-09 | fail→fail | 7,624 | 3,642 | -52% | 1 | 1 | 0% | 1,276 | 611 | -52% | 0 | 0 | — |
case-05 | pass→pass | 3,882 | 1,765 | -55% | 1 | 1 | 0% | 552 | 852 | +54% | 0 | 0 | — |
case-06 | fail→pass | 10,104 | 7,148 | -29% | 1 | 1 | 0% | 1,778 | 1,616 | -9% | 0 | 0 | — |
case-07 | fail→fail | 5,479 | 2,982 | -46% | 1 | 1 | 0% | 835 | 796 | -5% | 0 | 0 | — |
case-08 | fail→pass | 8,378 | 2,197 | -74% | 1 | 1 | 0% | 1,319 | 975 | -26% | 0 | 0 | — |
case-10 | fail→pass | 3,884 | 8,429 | +117% | 1 | 1 | 0% | 763 | 2,288 | +200% | 0 | 0 | — |
case-11 | pass→pass | 5,511 | 5,840 | +6% | 1 | 1 | 0% | 1,015 | 1,749 | +72% | 0 | 0 | — |
case-12 | fail→pass | 5,578 | 8,622 | +55% | 1 | 1 | 0% | 755 | 2,315 | +207% | 0 | 0 | — |
case-13 | fail→fail | 6,813 | 4,773 | -30% | 1 | 1 | 0% | 1,203 | 605 | -50% | 0 | 0 | — |
case-14 | fail→fail | 3,582 | 2,741 | -23% | 1 | 1 | 0% | 534 | 590 | +10% | 0 | 0 | — |
case-15 | fail→fail | 7,130 | 3,772 | -47% | 1 | 1 | 0% | 1,330 | 736 | -45% | 0 | 0 | — |
case-16 | fail→fail | 4,179 | 2,552 | -39% | 1 | 1 | 0% | 688 | 790 | +15% | 0 | 0 | — |
case-17 | fail→fail | 8,621 | 3,610 | -58% | 1 | 1 | 0% | 904 | 809 | -11% | 0 | 0 | — |
case-18 | pass→fail | 6,853 | 5,365 | -22% | 1 | 1 | 0% | 1,135 | 1,031 | -9% | 0 | 0 | — |
case-19 | pass→pass | 3,173 | 6,866 | +116% | 1 | 1 | 0% | 508 | 1,100 | +117% | 0 | 0 | — |
case-20 | pass→pass | 5,929 | 3,026 | -49% | 1 | 1 | 0% | 978 | 1,074 | +10% | 0 | 0 | — |
case-21 | fail→fail | 5,762 | 4,619 | -20% | 1 | 1 | 0% | 965 | 596 | -38% | 0 | 0 | — |
case-22 | fail→fail | 10,066 | 5,935 | -41% | 1 | 1 | 0% | 1,424 | 886 | -38% | 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 9 counted toward the lift figure. The other 13 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 +14 percentage points is the difference between those two pass rates over the 9 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.