Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Canonical watch/monitor pack for meaningful change detection across web sources, connectors, repo activity, or triggers, with false-positive suppression and alert summaries.
.claude/skills/mkurman-watch-monitor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -9% | 0% |
I want to define a watch that only notifies me on meaningful changes, whether the source is a webpage, connector-backed resource, repo activity, or trigger event.
watch_source: webpage / connector resource / repo / webhook familythreshold: significance threshold or rule setsuppression_rules: optional false-positive suppression criteriadelivery_channel: in-app or chat channelRun the Watch/Monitor pack every hour for the chosen source and notify only when the threshold is exceeded.
Use as the target logic behind a packaged trigger that ingests webhook events and converts them into operator-safe summaries.
Set up the Watch/Monitor pack for repo activity with stale-noise suppression and mobile-safe alerts.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→pass | 13,679 | 2,219 | -84% | 1 | 1 | 0% | 1,247 | 764 | -39% | 0 | 0 | — |
case-01 | fail→pass | 20,059 | 11,405 | -43% | 1 | 1 | 0% | 3,717 | 2,313 | -38% | 0 | 0 | — |
case-02 | pass→pass | 12,852 | 5,708 | -56% | 1 | 1 | 0% | 1,963 | 1,460 | -26% | 0 | 0 | — |
case-03 | pass→pass | 8,525 | 3,499 | -59% | 1 | 1 | 0% | 1,216 | 915 | -25% | 0 | 0 | — |
case-04 | pass→pass | 10,947 | 3,965 | -64% | 1 | 1 | 0% | 1,649 | 1,037 | -37% | 0 | 0 | — |
case-05 | fail→pass | 5,047 | 1,549 | -69% | 1 | 1 | 0% | 656 | 636 | -3% | 0 | 0 | — |
case-06 | fail→pass | 7,054 | 2,059 | -71% | 1 | 1 | 0% | 1,099 | 739 | -33% | 0 | 0 | — |
case-07 | fail→pass | 17,435 | 14,170 | -19% | 1 | 1 | 0% | 3,146 | 2,796 | -11% | 0 | 0 | — |
case-08 | pass→pass | 12,124 | 6,264 | -48% | 1 | 1 | 0% | 1,962 | 1,353 | -31% | 0 | 0 | — |
case-09 | pass→pass | 18,913 | 5,623 | -70% | 1 | 1 | 0% | 1,987 | 1,250 | -37% | 0 | 0 | — |
case-10 | pass→pass | 16,449 | 14,546 | -12% | 1 | 1 | 0% | 2,697 | 2,920 | +8% | 0 | 0 | — |
case-11 | fail→pass | 15,800 | 10,971 | -31% | 1 | 1 | 0% | 2,449 | 2,239 | -9% | 0 | 0 | — |
case-12 | fail→pass | 9,599 | 1,889 | -80% | 1 | 1 | 0% | 1,480 | 621 | -58% | 0 | 0 | — |
case-13 | fail→pass | 11,294 | 1,616 | -86% | 1 | 1 | 0% | 1,791 | 640 | -64% | 0 | 0 | — |
case-14 | fail→pass | 7,256 | 3,202 | -56% | 1 | 1 | 0% | 1,125 | 905 | -20% | 0 | 0 | — |
case-15 | pass→pass | 10,181 | 2,402 | -76% | 1 | 1 | 0% | 1,795 | 687 | -62% | 0 | 0 | — |
case-16 | pass→pass | 12,417 | 5,130 | -59% | 1 | 1 | 0% | 1,800 | 1,115 | -38% | 0 | 0 | — |
case-18 | pass→pass | 6,927 | 2,266 | -67% | 1 | 1 | 0% | 962 | 749 | -22% | 0 | 0 | — |
case-19 | fail→pass | 8,736 | 1,823 | -79% | 1 | 1 | 0% | 1,192 | 710 | -40% | 0 | 0 | — |
case-20 | pass→pass | 10,661 | 2,481 | -77% | 1 | 1 | 0% | 1,717 | 720 | -58% | 0 | 0 | — |
case-21 | pass→pass | 13,805 | 4,128 | -70% | 1 | 1 | 0% | 1,936 | 1,023 | -47% | 0 | 0 | — |
case-22 | fail→pass | 13,982 | 10,263 | -27% | 1 | 1 | 0% | 2,097 | 1,816 | -13% | 0 | 0 | — |
case-23 | pass→pass | 8,860 | 8,618 | -3% | 1 | 1 | 0% | 1,674 | 2,061 | +23% | 0 | 0 | — |
case-24 | pass→pass | 5,862 | 6,366 | +9% | 1 | 1 | 0% | 1,191 | 1,579 | +33% | 0 | 0 | — |
case-25 | pass→pass | 12,364 | 11,506 | -7% | 1 | 1 | 0% | 2,427 | 2,814 | +16% | 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. 25 cases were attempted. The headline lift of +40 percentage points is the difference between those two pass rates over the 25 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.