Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-perspective consensus planning with Planner/Architect/Critic loop. Use when high-stakes decisions need RALPLAN-DR structured deliberation (auth, migrations, public APIs, irreversible changes). Pairs with @compass-planner + @apex-architect + @raven-critic.
.claude/skills/evolution-foundation-dev-ralplan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 6% | 0% |
Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.
Consensus mode planning: a Planner/Architect/Critic loop that produces a plan with explicit principles, decision drivers, viable options, and an ADR. Used for high-stakes or irreversible work.
dev-plan --consensus would normally applydev-plan standard modedev-autopilotdev-plan direct mode@compass-planner)@apex-architect)@raven-critic)@compass-planner)--deliberate): + pre-mortem + expanded test plan + principle violation flagsTrigger deliberate mode when: auth/security touched, data migration, destructive/irreversible change, production incident response, compliance/PII implications, public API breakage.
Final plan saved to workspace/development/plans/[C]ralplan-{name}-{date}.md with ADR section.
@compass-planner (drives the loop)@apex-architect (architecture review)@raven-critic (adversarial review)dev-plan (which can chain into ralplan with --consensus)dev-autopilot (skips its own Phase 0+1 if a ralplan plan already exists)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 38,083 | 27,778 | -27% | 1 | 1 | 0% | 6,227 | 4,935 | -21% | 0 | 0 | — |
case-02 | fail→pass | 24,157 | 28,492 | +18% | 1 | 1 | 0% | 3,797 | 5,270 | +39% | 0 | 0 | — |
case-03 | fail→pass | 35,299 | 31,090 | -12% | 1 | 1 | 0% | 5,537 | 5,463 | -1% | 0 | 0 | — |
case-04 | pass→pass | 6,353 | 4,193 | -34% | 1 | 1 | 0% | 907 | 1,306 | +44% | 0 | 0 | — |
case-05 | pass→pass | 8,651 | 3,362 | -61% | 1 | 1 | 0% | 1,235 | 1,182 | -4% | 0 | 0 | — |
case-06 | pass→pass | 7,693 | 4,185 | -46% | 1 | 1 | 0% | 1,151 | 1,294 | +12% | 0 | 0 | — |
case-07 | fail→pass | 17,132 | 9,897 | -42% | 1 | 1 | 0% | 2,653 | 2,385 | -10% | 0 | 0 | — |
case-08 | pass→pass | 34,651 | 26,990 | -22% | 1 | 1 | 0% | 2,592 | 4,944 | +91% | 0 | 0 | — |
case-09 | fail→pass | 11,145 | 6,845 | -39% | 1 | 1 | 0% | 1,705 | 1,815 | +6% | 0 | 0 | — |
case-10 | pass→fail | 11,165 | 3,909 | -65% | 1 | 1 | 0% | 1,656 | 1,268 | -23% | 0 | 0 | — |
case-11 | fail→fail | 7,931 | 3,054 | -61% | 1 | 1 | 0% | 1,234 | 1,125 | -9% | 0 | 0 | — |
case-12 | fail→pass | 9,854 | 2,709 | -73% | 1 | 1 | 0% | 1,541 | 1,050 | -32% | 0 | 0 | — |
case-13 | fail→pass | 8,877 | 2,804 | -68% | 1 | 1 | 0% | 1,449 | 1,106 | -24% | 0 | 0 | — |
case-14 | pass→pass | 7,370 | 3,996 | -46% | 1 | 1 | 0% | 1,126 | 1,341 | +19% | 0 | 0 | — |
case-15 | fail→pass | 11,278 | 3,220 | -71% | 1 | 1 | 0% | 1,909 | 1,122 | -41% | 0 | 0 | — |
case-16 | fail→pass | 20,377 | 2,882 | -86% | 1 | 1 | 0% | 2,597 | 1,086 | -58% | 0 | 0 | — |
case-17 | pass→pass | 7,153 | 3,424 | -52% | 1 | 1 | 0% | 1,143 | 1,121 | -2% | 0 | 0 | — |
case-18 | fail→pass | 13,974 | 7,026 | -50% | 1 | 1 | 0% | 2,139 | 1,745 | -18% | 0 | 0 | — |
case-19 | fail→pass | 17,238 | 31,659 | +84% | 1 | 1 | 0% | 2,662 | 5,298 | +99% | 0 | 0 | — |
case-20 | fail→fail | 7,765 | 1,358 | -83% | 1 | 1 | 0% | 1,129 | 828 | -27% | 0 | 0 | — |
case-21 | fail→fail | 12,131 | 3,000 | -75% | 1 | 1 | 0% | 1,623 | 1,094 | -33% | 0 | 0 | — |
case-22 | fail→pass | 9,271 | 1,950 | -79% | 1 | 1 | 0% | 1,380 | 913 | -34% | 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 +50 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.