Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Summarize newsletter/email feed messages from the last N hours and propose follow-ups.
.claude/skills/cowork-os-usecase-newsletter-digest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 191% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 432% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 177% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 219% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 165% | 0% |
Summarize newsletter/email feed messages from the last N hours and propose follow-ups.
| Name | Type | Required | Description | |---|---|---|---| | channel | select | Yes | Where the newsletter feed arrives | | chat_hint | string | Yes | Name/keyword to identify the chat in channel_list_chats results | | since | string | No | Time window (e.g., "24h") | | message_limit | number | No | How many messages to fetch | | output_format | select | No | How to format the digest |
../usecase-newsletter-digest.json.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 18,632 | 9,042 | -51% | 1 | 1 | 0% | 2,879 | 1,828 | -37% | 0 | 0 | — |
case-01 | fail→fail | 2,665 | 7,063 | +165% | 1 | 1 | 0% | 398 | 1,529 | +284% | 0 | 0 | — |
case-02 | fail→pass | 4,810 | 9,859 | +105% | 1 | 1 | 0% | 642 | 1,871 | +191% | 0 | 0 | — |
case-03 | fail→pass | 2,361 | 12,044 | +410% | 1 | 1 | 0% | 366 | 1,947 | +432% | 0 | 0 | — |
case-04 | pass→pass | 18,029 | 2,912 | -84% | 1 | 1 | 0% | 2,874 | 837 | -71% | 0 | 0 | — |
case-05 | pass→pass | 19,998 | 9,012 | -55% | 1 | 1 | 0% | 2,925 | 1,702 | -42% | 0 | 0 | — |
case-07 | fail→fail | 5,065 | 10,243 | +102% | 1 | 1 | 0% | 783 | 1,752 | +124% | 0 | 0 | — |
case-08 | pass→pass | 3,066 | 6,733 | +120% | 1 | 1 | 0% | 425 | 1,416 | +233% | 0 | 0 | — |
case-09 | fail→pass | 3,998 | 7,628 | +91% | 1 | 1 | 0% | 541 | 1,496 | +177% | 0 | 0 | — |
case-10 | pass→pass | 9,515 | 9,311 | -2% | 1 | 1 | 0% | 1,473 | 1,899 | +29% | 0 | 0 | — |
case-11 | fail→pass | 3,140 | 6,296 | +101% | 1 | 1 | 0% | 417 | 1,330 | +219% | 0 | 0 | — |
case-12 | fail→pass | 3,656 | 9,000 | +146% | 1 | 1 | 0% | 514 | 1,364 | +165% | 0 | 0 | — |
case-13 | fail→fail | 5,517 | 7,589 | +38% | 1 | 1 | 0% | 791 | 1,478 | +87% | 0 | 0 | — |
case-14 | fail→pass | 5,523 | 10,643 | +93% | 1 | 1 | 0% | 906 | 2,069 | +128% | 0 | 0 | — |
case-15 | fail→fail | 10,794 | 11,821 | +10% | 1 | 1 | 0% | 709 | 1,930 | +172% | 0 | 0 | — |
case-16 | fail→pass | 2,652 | 8,103 | +206% | 1 | 1 | 0% | 409 | 1,673 | +309% | 0 | 0 | — |
case-17 | fail→pass | 4,353 | 13,037 | +199% | 1 | 1 | 0% | 645 | 2,336 | +262% | 0 | 0 | — |
case-18 | fail→pass | 5,898 | 11,290 | +91% | 1 | 1 | 0% | 465 | 1,583 | +240% | 0 | 0 | — |
case-19 | fail→pass | 5,105 | 10,943 | +114% | 1 | 1 | 0% | 777 | 2,062 | +165% | 0 | 0 | — |
case-20 | fail→pass | 5,700 | 11,289 | +98% | 1 | 1 | 0% | 784 | 2,170 | +177% | 0 | 0 | — |
case-21 | fail→pass | 4,801 | 12,692 | +164% | 1 | 1 | 0% | 683 | 1,811 | +165% | 0 | 0 | — |
case-22 | fail→pass | 3,955 | 7,760 | +96% | 1 | 1 | 0% | 553 | 1,503 | +172% | 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 20 counted toward the lift figure. The other 2 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 +59 percentage points is the difference between those two pass rates over the 20 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.