Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Watches specific web pages and reports the day anything changes — pricing, job postings, product updates. The user adds a page with "watch: [URL] — [label]"; the agent snapshots it to memory and re-checks daily, alerting on Telegram with exactly what moved (old price → new price, added/removed plans, new job postings).
.claude/skills/nearai-competitor-page-watcher/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -48% | 0% |
You watch web pages and tell the user the day anything meaningful changes — pricing, job postings, product updates.
HEARTBEAT_OK and stop; never send a "nothing changed" message.When the user says watch: [URL] — [what it is] (e.g. watch: https://competitor.com/pricing — their pricing page):
http tool.competitors/pages.md with memory_read.memory_write.Create a routine that runs every day at 8:00 AM:
competitors/pages.md.HEARTBEAT_OK and stop.Alert format: 🔍 Competitor Change Detected — date] 📄 Competitor pricing page]
📄 Competitor careers page]
💡 What this might signal: one line per change]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→pass | 10,088 | 2,084 | -79% | 1 | 1 | 0% | 1,695 | 956 | -44% | 0 | 0 | — |
case-01 | fail→fail | 4,564 | 4,540 | -1% | 1 | 1 | 0% | 679 | 890 | +31% | 0 | 0 | — |
case-02 | fail→fail | 2,953 | 5,288 | +79% | 1 | 1 | 0% | 400 | 819 | +105% | 0 | 0 | — |
case-03 | fail→fail | 6,560 | 4,056 | -38% | 1 | 1 | 0% | 1,071 | 776 | -28% | 0 | 0 | — |
case-04 | fail→pass | 12,689 | 3,908 | -69% | 1 | 1 | 0% | 2,484 | 1,198 | -52% | 0 | 0 | — |
case-05 | pass→pass | 11,594 | 9,087 | -22% | 1 | 1 | 0% | 679 | 1,022 | +51% | 0 | 0 | — |
case-06 | fail→pass | 14,479 | 4,694 | -68% | 1 | 1 | 0% | 2,825 | 1,459 | -48% | 0 | 0 | — |
case-08 | pass→pass | 11,568 | 3,504 | -70% | 1 | 1 | 0% | 2,012 | 1,265 | -37% | 0 | 0 | — |
case-09 | fail→pass | 4,635 | 1,656 | -64% | 1 | 1 | 0% | 784 | 891 | +14% | 0 | 0 | — |
case-10 | fail→pass | 13,823 | 1,663 | -88% | 1 | 1 | 0% | 1,682 | 879 | -48% | 0 | 0 | — |
case-11 | pass→pass | 11,009 | 2,617 | -76% | 1 | 1 | 0% | 1,784 | 999 | -44% | 0 | 0 | — |
case-12 | fail→fail | 3,142 | 4,551 | +45% | 1 | 1 | 0% | 448 | 906 | +102% | 0 | 0 | — |
case-13 | fail→pass | 17,410 | 2,277 | -87% | 1 | 1 | 0% | 2,427 | 938 | -61% | 0 | 0 | — |
case-14 | fail→pass | 7,598 | 1,341 | -82% | 1 | 1 | 0% | 1,316 | 786 | -40% | 0 | 0 | — |
case-15 | pass→pass | 9,685 | 3,616 | -63% | 1 | 1 | 0% | 1,681 | 1,134 | -33% | 0 | 0 | — |
case-16 | pass→pass | 8,450 | 2,810 | -67% | 1 | 1 | 0% | 1,501 | 1,136 | -24% | 0 | 0 | — |
case-17 | fail→fail | 13,497 | 1,750 | -87% | 1 | 1 | 0% | 2,482 | 910 | -63% | 0 | 0 | — |
case-18 | fail→fail | 6,499 | 4,288 | -34% | 1 | 1 | 0% | 1,161 | 908 | -22% | 0 | 0 | — |
case-19 | fail→fail | 9,508 | 3,080 | -68% | 1 | 1 | 0% | 1,650 | 1,192 | -28% | 0 | 0 | — |
case-20 | fail→pass | 4,800 | 1,654 | -66% | 1 | 1 | 0% | 712 | 875 | +23% | 0 | 0 | — |
case-21 | pass→fail | 4,770 | 1,517 | -68% | 1 | 1 | 0% | 756 | 860 | +14% | 0 | 0 | — |
case-22 | fail→pass | 10,865 | 3,562 | -67% | 1 | 1 | 0% | 1,880 | 857 | -54% | 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 17 counted toward the lift figure. The other 5 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 +36 percentage points is the difference between those two pass rates over the 17 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.