Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Request cross-AI peer review of phase plans from external AI CLIs
.claude/skills/davepoon-gsd-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -59% | 0% |
<objective> Invoke external AI CLIs (Gemini, Claude, Codex, OpenCode, Qwen Code, Cursor) to independently review phase plans. Produces a structured REVIEWS.md with per-reviewer feedback that can be fed back into planning via /gsd:plan-phase --reviews.
Flow: Detect CLIs → Build review prompt → Invoke each CLI → Collect responses → Write REVIEWS.md </objective>
<execution_context> @${CLAUDE_PLUGIN_ROOT}/workflows/review.md </execution_context>
<context> Phase number: extracted from $ARGUMENTS (required)
Flags:
--gemini — Include Gemini CLI review--claude — Include Claude CLI review (uses separate session)--codex — Include Codex CLI review--opencode — Include OpenCode review (uses model from user's OpenCode config)--qwen — Include Qwen Code review (Alibaba Qwen models)--cursor — Include Cursor agent review--all — Include all available CLIs</context>
<process> Execute the review workflow from @${CLAUDE_PLUGIN_ROOT}/workflows/review.md end-to-end. </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 | 6,333 | 19,180 | +203% | 1 | 1 | 0% | 1,132 | 3,317 | +193% | 0 | 0 | — |
case-02 | fail→fail | 4,670 | 6,659 | +43% | 1 | 1 | 0% | 184 | 568 | +209% | 0 | 0 | — |
case-03 | fail→fail | 4,948 | 5,583 | +13% | 1 | 1 | 0% | 251 | 731 | +191% | 0 | 0 | — |
case-04 | fail→pass | 9,781 | 2,384 | -76% | 1 | 1 | 0% | 1,648 | 696 | -58% | 0 | 0 | — |
case-05 | fail→pass | 10,910 | 3,600 | -67% | 1 | 1 | 0% | 1,796 | 905 | -50% | 0 | 0 | — |
case-06 | fail→pass | 9,426 | 2,541 | -73% | 1 | 1 | 0% | 1,697 | 708 | -58% | 0 | 0 | — |
case-07 | fail→pass | 6,731 | 2,458 | -63% | 1 | 1 | 0% | 1,232 | 695 | -44% | 0 | 0 | — |
case-08 | fail→pass | 8,517 | 1,990 | -77% | 1 | 1 | 0% | 1,452 | 595 | -59% | 0 | 0 | — |
case-09 | fail→pass | 10,376 | 3,111 | -70% | 1 | 1 | 0% | 1,746 | 845 | -52% | 0 | 0 | — |
case-10 | pass→pass | 5,981 | 1,868 | -69% | 1 | 1 | 0% | 1,063 | 603 | -43% | 0 | 0 | — |
case-11 | fail→pass | 10,340 | 3,292 | -68% | 1 | 1 | 0% | 1,769 | 873 | -51% | 0 | 0 | — |
case-12 | fail→pass | 9,445 | 2,924 | -69% | 1 | 1 | 0% | 1,482 | 724 | -51% | 0 | 0 | — |
case-13 | fail→pass | 9,942 | 3,998 | -60% | 1 | 1 | 0% | 1,662 | 960 | -42% | 0 | 0 | — |
case-14 | fail→pass | 11,362 | 2,196 | -81% | 1 | 1 | 0% | 1,790 | 677 | -62% | 0 | 0 | — |
case-15 | fail→pass | 12,507 | 5,366 | -57% | 1 | 1 | 0% | 2,004 | 1,174 | -41% | 0 | 0 | — |
case-16 | pass→pass | 7,348 | 2,330 | -68% | 1 | 1 | 0% | 1,278 | 678 | -47% | 0 | 0 | — |
case-17 | pass→pass | 4,298 | 1,325 | -69% | 1 | 1 | 0% | 726 | 423 | -42% | 0 | 0 | — |
case-18 | fail→pass | 11,513 | 5,954 | -48% | 1 | 1 | 0% | 2,121 | 1,422 | -33% | 0 | 0 | — |
case-19 | fail→pass | 12,082 | 2,475 | -80% | 1 | 1 | 0% | 2,211 | 706 | -68% | 0 | 0 | — |
case-20 | pass→fail | 16,275 | 4,934 | -70% | 1 | 1 | 0% | 2,915 | 559 | -81% | 0 | 0 | — |
case-21 | fail→fail | 4,333 | 5,518 | +27% | 1 | 1 | 0% | 671 | 617 | -8% | 0 | 0 | — |
case-22 | fail→fail | 5,085 | 3,647 | -28% | 1 | 1 | 0% | 838 | 538 | -36% | 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 16 counted toward the lift figure. The other 6 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 +55 percentage points is the difference between those two pass rates over the 16 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.