Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Make a recurring brief report what changed since the last edition instead of restating everything. Use when a weekly or monthly report keeps repeating itself, when setting up a scheduled monitor or digest, or when asked to make a recurring update delta-aware. Produces a changes-first brief plus the state record the next run will diff against.
.claude/skills/mohitagw15856-delta-briefing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 44% | 0% |
The failure mode of every recurring report is that edition 6 reads like edition 5. This skill structures a recurring brief around the delta: read the last edition's state, diff the world against it, lead with what changed, and save state for the next run.
Ask for (if not already provided):
TL;DR: 1-2 sentences: the most consequential delta, or "no material changes"]
New since last edition
Changed
Resolved
Still watching (one line each)
<details> <summary>State record (for the next run)</summary>
json{ "edition": n, "date": "YYYY-MM-DD", "sources": ["..."], "items": [ { "id": "...", "status": "...", "note": "..." } ] }
</details>
If nothing material changed: say exactly that in three lines — TL;DR ("no material changes"), what was checked, next edition date. Do not pad.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→pass | 17,366 | 12,845 | -26% | 1 | 1 | 0% | 2,667 | 2,989 | +12% | 0 | 0 | — |
case-01 | fail→fail | 19,840 | 4,636 | -77% | 1 | 1 | 0% | 3,212 | 1,559 | -51% | 0 | 0 | — |
case-02 | fail→fail | 19,368 | 5,845 | -70% | 1 | 1 | 0% | 3,144 | 1,885 | -40% | 0 | 0 | — |
case-03 | fail→pass | 19,398 | 11,614 | -40% | 1 | 1 | 0% | 3,563 | 3,117 | -13% | 0 | 0 | — |
case-04 | fail→pass | 8,290 | 4,338 | -48% | 1 | 1 | 0% | 1,420 | 1,850 | +30% | 0 | 0 | — |
case-05 | pass→pass | 10,328 | 6,106 | -41% | 1 | 1 | 0% | 1,604 | 2,142 | +34% | 0 | 0 | — |
case-06 | pass→pass | 4,866 | 6,512 | +34% | 1 | 1 | 0% | 806 | 2,206 | +174% | 0 | 0 | — |
case-07 | fail→pass | 9,094 | 5,534 | -39% | 1 | 1 | 0% | 1,430 | 2,039 | +43% | 0 | 0 | — |
case-14 | pass→pass | 12,485 | 10,624 | -15% | 1 | 1 | 0% | 1,949 | 2,596 | +33% | 0 | 0 | — |
case-08 | fail→pass | 7,135 | 6,478 | -9% | 1 | 1 | 0% | 1,473 | 2,128 | +44% | 0 | 0 | — |
case-09 | fail→fail | 14,842 | 4,913 | -67% | 1 | 1 | 0% | 2,835 | 1,709 | -40% | 0 | 0 | — |
case-10 | fail→pass | 10,741 | 6,072 | -43% | 1 | 1 | 0% | 1,719 | 1,948 | +13% | 0 | 0 | — |
case-11 | pass→pass | 16,463 | 9,104 | -45% | 1 | 1 | 0% | 2,935 | 2,317 | -21% | 0 | 0 | — |
case-12 | fail→pass | 8,493 | 2,924 | -66% | 1 | 1 | 0% | 1,258 | 1,351 | +7% | 0 | 0 | — |
case-15 | pass→pass | 13,537 | 8,026 | -41% | 1 | 1 | 0% | 2,085 | 2,210 | +6% | 0 | 0 | — |
case-16 | pass→pass | 5,757 | 6,792 | +18% | 1 | 1 | 0% | 1,192 | 2,210 | +85% | 0 | 0 | — |
case-17 | pass→pass | 6,739 | 4,325 | -36% | 1 | 1 | 0% | 1,255 | 1,679 | +34% | 0 | 0 | — |
case-18 | pass→pass | 7,094 | 6,950 | -2% | 1 | 1 | 0% | 1,599 | 2,119 | +33% | 0 | 0 | — |
case-19 | pass→fail | 30,545 | 15,315 | -50% | 1 | 1 | 0% | 6,186 | 3,847 | -38% | 0 | 0 | — |
case-20 | pass→pass | 19,991 | 16,990 | -15% | 1 | 1 | 0% | 4,304 | 4,612 | +7% | 0 | 0 | — |
case-21 | pass→pass | 17,295 | 16,761 | -3% | 1 | 1 | 0% | 2,920 | 3,643 | +25% | 0 | 0 | — |
case-22 | pass→pass | 4,472 | 5,543 | +24% | 1 | 1 | 0% | 722 | 2,022 | +180% | 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. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.