Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Find the subscriptions you forgot you pay for — a tool-using agent audits statements and inboxes, prices the waste annually, and preps (never executes) the cancellations. Use when asked to audit my subscriptions, find recurring charges, what am I paying for, or help me cancel unused services. Produces the subscription inventory with keep/cancel/downgrade verdicts, the annual-waste number, and approval-gated cancellation prep.
.claude/skills/mohitagw15856-subscription-auditor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 191% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-06 | ✓→✓ | = Same ✓ | 36% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 222% | 0% |
Every card statement hides a museum of past enthusiasms billing monthly. This skill runs the audit an accountant would: inventory every recurring charge, price the waste annually (monthly numbers anesthetize), verdict each one — and stop exactly at the line where money moves. Cancellation is prepared, never performed.
Ask for these if not provided:
The number: $n]/year in cancel-verdict subscriptions. | Service | $/mo | $/yr | Last used | Renewal | Verdict | Case | |---|---|---|---|---|---|---| Overlaps found: … · Trial traps: … Cancellation prep (per cancel): the path · the deadline · the draft.
For tool-using agents with statement/inbox read access and (optionally) browser control. Without tools, run on user-pasted statements. Rules per SKILLSPEC.md §5.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 5,069 | 13,586 | +168% | 1 | 1 | 0% | 853 | 2,485 | +191% | 0 | 0 | — |
case-02 | fail→fail | 31,843 | 20,999 | -34% | 1 | 1 | 0% | 4,740 | 3,692 | -22% | 0 | 0 | — |
case-03 | fail→fail | 17,018 | 14,930 | -12% | 1 | 1 | 0% | 887 | 2,222 | +151% | 0 | 0 | — |
case-04 | fail→pass | 14,482 | 11,454 | -21% | 1 | 1 | 0% | 1,520 | 2,235 | +47% | 0 | 0 | — |
case-05 | fail→pass | 17,701 | 8,028 | -55% | 1 | 1 | 0% | 1,855 | 2,213 | +19% | 0 | 0 | — |
case-06 | pass→pass | 9,112 | 10,371 | +14% | 1 | 1 | 0% | 1,504 | 2,039 | +36% | 0 | 0 | — |
case-07 | pass→pass | 9,352 | 8,059 | -14% | 1 | 1 | 0% | 725 | 2,331 | +222% | 0 | 0 | — |
case-08 | pass→pass | 16,613 | 9,608 | -42% | 1 | 1 | 0% | 1,928 | 1,646 | -15% | 0 | 0 | — |
case-09 | pass→pass | 13,858 | 9,507 | -31% | 1 | 1 | 0% | 1,384 | 1,663 | +20% | 0 | 0 | — |
case-10 | pass→pass | 13,792 | 14,874 | +8% | 1 | 1 | 0% | 1,508 | 2,407 | +60% | 0 | 0 | — |
case-11 | fail→fail | 14,742 | 17,807 | +21% | 1 | 1 | 0% | 2,051 | 4,005 | +95% | 0 | 0 | — |
case-12 | pass→pass | 14,074 | 3,785 | -73% | 1 | 1 | 0% | 1,134 | 1,590 | +40% | 0 | 0 | — |
case-13 | pass→pass | 9,975 | 16,160 | +62% | 1 | 1 | 0% | 1,431 | 2,513 | +76% | 0 | 0 | — |
case-14 | pass→pass | 12,783 | 4,607 | -64% | 1 | 1 | 0% | 956 | 1,717 | +80% | 0 | 0 | — |
case-15 | pass→pass | 15,032 | 10,695 | -29% | 1 | 1 | 0% | 1,577 | 1,853 | +18% | 0 | 0 | — |
case-16 | pass→pass | 7,914 | 11,219 | +42% | 1 | 1 | 0% | 1,298 | 2,031 | +56% | 0 | 0 | — |
case-17 | pass→pass | 14,885 | 12,806 | -14% | 1 | 1 | 0% | 1,499 | 2,271 | +52% | 0 | 0 | — |
case-18 | pass→pass | 7,376 | 7,390 | +0% | 1 | 1 | 0% | 914 | 2,255 | +147% | 0 | 0 | — |
case-19 | fail→fail | 14,014 | 8,625 | -38% | 1 | 1 | 0% | 1,337 | 2,455 | +84% | 0 | 0 | — |
case-20 | pass→pass | 13,552 | 13,444 | -1% | 1 | 1 | 0% | 2,195 | 3,231 | +47% | 0 | 0 | — |
case-21 | pass→pass | 14,107 | 19,978 | +42% | 1 | 1 | 0% | 2,278 | 3,286 | +44% | 0 | 0 | — |
case-22 | pass→pass | 16,049 | 14,728 | -8% | 1 | 1 | 0% | 2,403 | 3,235 | +35% | 0 | 0 | — |
case-23 | pass→pass | 14,086 | 11,877 | -16% | 1 | 1 | 0% | 2,291 | 2,848 | +24% | 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. 23 cases were attempted. The headline lift of +13 percentage points is the difference between those two pass rates over the 23 comparable cases.
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.