Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Research and write professional newsletters.
.claude/skills/hezaohezao-newsletter-generation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-09 | ✓→✗ | ▼ Worse | -8% | 0% |
| case-18 | ✓→✗ | ▼ Worse | -26% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -54% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 143% | 0% |
Generates professional, well-researched newsletters combining curated content from multiple sources with original analysis and commentary. Follows modern newsletter best practices (Morning Brew, The Hustle, TLDR, Benedict Evans).
Output is a complete, ready-to-publish newsletter in Markdown.
Identify:
Use web_search + browse_page to gather content:
browse_page on promising resultsStructure the newsletter:
markdown# [Newsletter Name] — [Date] ## This Week's Headline [1-2 sentence hook for the top story] --- ## Top Stories ### 1. [Headline] [2-3 sentence summary of the story] **Why it matters**: [1 sentence on significance] [Source](URL) ### 2. [Headline] [...] ### 3. [Headline] [...] --- ## Quick Hits - [Brief item with link](URL) - [Brief item with link](URL) - [Brief item with link](URL) --- ## Deep Dive: [Topic] [Longer analysis section on one key topic — 200-400 words. Original commentary, not just summary. Add your perspective on what this means for the audience.] --- ## Quote of the Week > "[Notable quote]" — [Attribution](URL) --- ## Coming Up Next Week [1-2 sentence preview of expected developments] --- *Curated by [Agent Name]. [Subscribe](link) | [Archive](link)*
Save to .poirot/outputs/newsletter-{topic}-{YYYYMMDD}.md via write_file. Present via present_files.
| Format | Items | Length | Cadence | |--------|-------|--------|---------| | Daily Digest | 3-5 | 300-500 words | Daily | | Weekly Roundup | 5-10 | 800-1500 words | Weekly | | Deep-Dive | 1-3 | 1500-3000 words | Irregular | | Industry Briefing | 5-8 | 600-1200 words | Weekly/Biweekly |
window).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,022 | 14,227 | -29% | 1 | 1 | 0% | 2,859 | 1,487 | -48% | 0 | 0 | — |
case-02 | fail→fail | 27,289 | 8,042 | -71% | 1 | 1 | 0% | 3,815 | 1,470 | -61% | 0 | 0 | — |
case-03 | fail→fail | 17,658 | 9,751 | -45% | 1 | 1 | 0% | 2,438 | 1,485 | -39% | 0 | 0 | — |
case-04 | pass→pass | 10,932 | 18,736 | +71% | 1 | 1 | 0% | 1,690 | 4,105 | +143% | 0 | 0 | — |
case-05 | pass→pass | 12,832 | 32,182 | +151% | 1 | 1 | 0% | 2,122 | 2,962 | +40% | 0 | 0 | — |
case-06 | fail→fail | 44,698 | 35,722 | -20% | 1 | 1 | 0% | 7,163 | 6,389 | -11% | 0 | 0 | — |
case-07 | pass→pass | 16,972 | 19,241 | +13% | 1 | 1 | 0% | 2,949 | 4,755 | +61% | 0 | 0 | — |
case-08 | fail→fail | 9,864 | 13,157 | +33% | 1 | 1 | 0% | 1,360 | 1,412 | +4% | 0 | 0 | — |
case-09 | pass→fail | 16,702 | 20,672 | +24% | 1 | 1 | 0% | 1,969 | 1,820 | -8% | 0 | 0 | — |
case-10 | pass→pass | 14,582 | 5,922 | -59% | 1 | 1 | 0% | 2,051 | 1,796 | -12% | 0 | 0 | — |
case-11 | fail→fail | 10,518 | 14,955 | +42% | 1 | 1 | 0% | 1,878 | 2,166 | +15% | 0 | 0 | — |
case-12 | fail→fail | 7,145 | 7,288 | +2% | 1 | 1 | 0% | 290 | 1,360 | +369% | 0 | 0 | — |
case-13 | fail→pass | 19,966 | 35,049 | +76% | 1 | 1 | 0% | 2,618 | 4,282 | +64% | 0 | 0 | — |
case-14 | fail→fail | 14,857 | 12,952 | -13% | 1 | 1 | 0% | 2,346 | 1,534 | -35% | 0 | 0 | — |
case-15 | fail→fail | 14,932 | 49,353 | +231% | 1 | 1 | 0% | 2,125 | 1,424 | -33% | 0 | 0 | — |
case-16 | fail→fail | 25,555 | 7,799 | -69% | 1 | 1 | 0% | 2,351 | 1,412 | -40% | 0 | 0 | — |
case-17 | fail→fail | 15,551 | 47,848 | +208% | 1 | 1 | 0% | 2,291 | 1,466 | -36% | 0 | 0 | — |
case-18 | pass→fail | 15,776 | 8,816 | -44% | 1 | 1 | 0% | 2,142 | 1,592 | -26% | 0 | 0 | — |
case-19 | fail→fail | 13,562 | 10,198 | -25% | 1 | 1 | 0% | 1,987 | 1,353 | -32% | 0 | 0 | — |
case-20 | fail→fail | 12,983 | 20,642 | +59% | 1 | 1 | 0% | 1,623 | 2,679 | +65% | 0 | 0 | — |
case-21 | pass→fail | 23,895 | 41,753 | +75% | 1 | 1 | 0% | 3,188 | 1,475 | -54% | 0 | 0 | — |
case-22 | fail→fail | 20,272 | 7,147 | -65% | 1 | 1 | 0% | 2,003 | 1,992 | -1% | 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 -9 percentage points is the difference between those two pass rates over the 9 comparable cases. 8 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.