Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →You produce a daily AI news digest. Your output is a single markdown file written to `content/research/{YYYY-MM-DD}-ai-news.md` and committed to a branch.
.claude/skills/owainlewis-ai-news-research-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -32% | 0% |
You produce a daily AI news digest. Your output is a single markdown file written to content/research/{YYYY-MM-DD}-ai-news.md and committed to a branch.
references/sources.md for the full listmarkdown--- date: {YYYY-MM-DD} generated_by: ai-news-research-agent --- # AI News Digest, {YYYY-MM-DD} ## Top stories - [Title](url) — one-line takeaway - [Title](url) — one-line takeaway - [Title](url) — one-line takeaway ## GitHub trending (AI agents) - [repo](url) — one-line description, stars ## Worth a closer look - [thing](url) — why it matters ## Quick links - [smaller item](url) - [smaller item](url)
Surface anything relevant to:
Skip:
Write to content/research/{YYYY-MM-DD}-ai-news.md in the configured working directory.
After writing the file:
agent/news-{YYYY-MM-DD}News digest for {YYYY-MM-DD}| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 10,690 | 3,201 | -70% | 1 | 1 | 0% | 1,516 | 930 | -39% | 0 | 0 | — |
case-01 | fail→fail | 6,992 | 7,230 | +3% | 1 | 1 | 0% | 1,161 | 701 | -40% | 0 | 0 | — |
case-02 | fail→fail | 15,538 | 11,014 | -29% | 1 | 1 | 0% | 2,371 | 1,488 | -37% | 0 | 0 | — |
case-03 | fail→fail | 13,587 | 6,790 | -50% | 1 | 1 | 0% | 2,031 | 851 | -58% | 0 | 0 | — |
case-04 | fail→pass | 5,933 | 1,848 | -69% | 1 | 1 | 0% | 922 | 774 | -16% | 0 | 0 | — |
case-05 | fail→pass | 7,375 | 2,346 | -68% | 1 | 1 | 0% | 1,175 | 848 | -28% | 0 | 0 | — |
case-06 | pass→pass | 3,917 | 3,376 | -14% | 1 | 1 | 0% | 595 | 1,007 | +69% | 0 | 0 | — |
case-07 | fail→pass | 4,998 | 4,606 | -8% | 1 | 1 | 0% | 763 | 1,213 | +59% | 0 | 0 | — |
case-08 | fail→pass | 13,101 | 4,863 | -63% | 1 | 1 | 0% | 1,764 | 1,202 | -32% | 0 | 0 | — |
case-10 | fail→pass | 4,695 | 8,053 | +72% | 1 | 1 | 0% | 621 | 1,633 | +163% | 0 | 0 | — |
case-11 | pass→pass | 9,930 | 3,019 | -70% | 1 | 1 | 0% | 1,308 | 920 | -30% | 0 | 0 | — |
case-12 | fail→pass | 11,728 | 3,470 | -70% | 1 | 1 | 0% | 788 | 1,048 | +33% | 0 | 0 | — |
case-13 | fail→pass | 6,676 | 4,859 | -27% | 1 | 1 | 0% | 918 | 1,215 | +32% | 0 | 0 | — |
case-14 | fail→fail | 10,747 | 1,598 | -85% | 1 | 1 | 0% | 1,586 | 677 | -57% | 0 | 0 | — |
case-15 | pass→pass | 13,687 | 2,474 | -82% | 1 | 1 | 0% | 2,266 | 812 | -64% | 0 | 0 | — |
case-16 | fail→pass | 10,545 | 4,525 | -57% | 1 | 1 | 0% | 1,489 | 1,177 | -21% | 0 | 0 | — |
case-17 | fail→pass | 3,528 | 2,160 | -39% | 1 | 1 | 0% | 570 | 863 | +51% | 0 | 0 | — |
case-18 | pass→pass | 8,846 | 4,133 | -53% | 1 | 1 | 0% | 1,288 | 1,120 | -13% | 0 | 0 | — |
case-19 | pass→fail | 36,771 | 7,314 | -80% | 1 | 1 | 0% | 6,169 | 1,587 | -74% | 0 | 0 | — |
case-20 | pass→fail | 14,357 | 5,241 | -63% | 1 | 1 | 0% | 2,100 | 1,196 | -43% | 0 | 0 | — |
case-21 | pass→pass | 5,463 | 5,187 | -5% | 1 | 1 | 0% | 982 | 1,192 | +21% | 0 | 0 | — |
case-22 | fail→fail | 3,704 | 1,694 | -54% | 1 | 1 | 0% | 477 | 713 | +49% | 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 18 counted toward the lift figure. The other 4 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 18 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.