Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review and promote backlog items to active milestone
.claude/skills/davepoon-gsd-review-backlog/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-14 | ✓→✗ | ▼ Worse | -25% | 0% |
<objective> Review all 999.x backlog items and optionally promote them into the active milestone sequence or remove stale entries. </objective>
<process>
bash ls -d .planning/phases/999* 2>/dev/null || echo "No backlog items found"
bash cat .planning/ROADMAP.md Show each backlog item with its description, any accumulated context (CONTEXT.md, RESEARCH.md), and creation date.
999.x-slug to {new_num}-slug:bash NEW_NUM=$(gsd-sdk query phase.add "${DESCRIPTION}" --raw)
## Backlog section to the active phase list(BACKLOG) marker**Depends on:** field## Backlog sectionbash gsd-sdk query commit "docs: review backlog — promoted N, removed M" .planning/ROADMAP.md
## 📋 Backlog Review Complete
Promoted: {list of promoted items with new phase numbers} Kept: {list of items remaining in backlog} Removed: {list of deleted items}
</process>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,431 | 4,224 | -34% | 1 | 1 | 0% | 1,060 | 758 | -28% | 0 | 0 | — |
case-02 | fail→fail | 5,648 | 4,166 | -26% | 1 | 1 | 0% | 889 | 713 | -20% | 0 | 0 | — |
case-03 | fail→fail | 11,044 | 4,434 | -60% | 1 | 1 | 0% | 2,336 | 725 | -69% | 0 | 0 | — |
case-04 | fail→fail | 7,482 | 2,985 | -60% | 1 | 1 | 0% | 1,407 | 797 | -43% | 0 | 0 | — |
case-05 | fail→fail | 9,322 | 5,089 | -45% | 1 | 1 | 0% | 1,583 | 777 | -51% | 0 | 0 | — |
case-06 | fail→fail | 9,624 | 7,407 | -23% | 1 | 1 | 0% | 1,557 | 946 | -39% | 0 | 0 | — |
case-07 | fail→fail | 8,633 | 6,675 | -23% | 1 | 1 | 0% | 1,393 | 1,034 | -26% | 0 | 0 | — |
case-08 | fail→pass | 8,229 | 2,619 | -68% | 1 | 1 | 0% | 1,396 | 935 | -33% | 0 | 0 | — |
case-09 | fail→fail | 7,546 | 2,826 | -63% | 1 | 1 | 0% | 1,186 | 939 | -21% | 0 | 0 | — |
case-10 | fail→fail | 8,792 | 7,056 | -20% | 1 | 1 | 0% | 1,423 | 1,004 | -29% | 0 | 0 | — |
case-11 | fail→fail | 12,381 | 3,396 | -73% | 1 | 1 | 0% | 2,009 | 876 | -56% | 0 | 0 | — |
case-12 | fail→pass | 6,441 | 5,209 | -19% | 1 | 1 | 0% | 1,110 | 748 | -33% | 0 | 0 | — |
case-13 | pass→pass | 10,198 | 1,809 | -82% | 1 | 1 | 0% | 1,728 | 759 | -56% | 0 | 0 | — |
case-14 | pass→fail | 7,839 | 3,038 | -61% | 1 | 1 | 0% | 1,343 | 1,006 | -25% | 0 | 0 | — |
case-15 | pass→pass | 8,724 | 2,583 | -70% | 1 | 1 | 0% | 1,603 | 951 | -41% | 0 | 0 | — |
case-16 | pass→pass | 4,971 | 3,418 | -31% | 1 | 1 | 0% | 841 | 1,093 | +30% | 0 | 0 | — |
case-17 | pass→pass | 10,029 | 3,071 | -69% | 1 | 1 | 0% | 1,559 | 904 | -42% | 0 | 0 | — |
case-18 | fail→pass | 8,428 | 3,056 | -64% | 1 | 1 | 0% | 1,517 | 930 | -39% | 0 | 0 | — |
case-19 | fail→pass | 10,577 | 2,796 | -74% | 1 | 1 | 0% | 1,826 | 960 | -47% | 0 | 0 | — |
case-20 | fail→fail | 4,889 | 4,334 | -11% | 1 | 1 | 0% | 293 | 760 | +159% | 0 | 0 | — |
case-21 | fail→fail | 4,762 | 5,064 | +6% | 1 | 1 | 0% | 328 | 800 | +144% | 0 | 0 | — |
case-22 | pass→fail | 4,958 | 4,696 | -5% | 1 | 1 | 0% | 800 | 771 | -4% | 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 14 counted toward the lift figure. The other 8 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 14 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.