Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Full autonomous execution from idea to working code. Use when user says "autopilot", "autonomous", "build me", "create me", "make me", "full auto" — orchestrates spec → plan → code → QA → validation across the engineering layer agents.
.claude/skills/evolution-foundation-dev-autopilot/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 55% | 0% |
Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.
Autopilot takes a brief product idea and autonomously handles the full lifecycle: requirements analysis, technical design, planning, parallel implementation, QA cycling, and multi-perspective validation. It produces working, verified code from a 2-3 line description.
dev-plan instead@bolt-executordev-plan --reviewMost non-trivial software tasks require coordinated phases: understanding requirements, designing a solution, implementing in parallel, testing, and validating quality. Autopilot orchestrates all of these phases automatically so the user can describe what they want and receive working code without managing each step.
Turn the user's idea into a detailed spec.
dev-deep-interview for Socratic clarification before expanding@echo-analyst (Opus) extracts requirements, @apex-architect (Opus) creates technical specificationworkspace/projects/specs/[C]autopilot-spec-{name}.mdCreate an implementation plan from the spec.
@compass-planner (Opus): Create plan (direct mode, no interview)@raven-critic (Opus): Validate planworkspace/projects/plans/[C]autopilot-plan-{name}.mdImplement the plan.
@bolt-executor (Sonnet): Standard tasks@apex-architect for design before re-delegatingCycle until all tests pass.
Multi-perspective review in parallel.
@apex-architect: Functional completeness@vault-security: Vulnerability check (when imported in EPIC 3)@lens-reviewer: Quality review@oath-verifier: Final evidence-based completion check against acceptance criteriaGood:
→ Specific domain, clear features, technology constraint. Enough context to expand.
→ Clear product concept with a specific feature.
Bad:
@hawk-debugger or @bolt-executor.dev-plan.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 27,624 | 5,177 | -81% | 1 | 1 | 0% | 6,193 | 1,814 | -71% | 0 | 0 | — |
case-02 | fail→fail | 30,472 | 5,477 | -82% | 1 | 1 | 0% | 6,198 | 1,424 | -77% | 0 | 0 | — |
case-03 | fail→fail | 27,245 | 13,644 | -50% | 1 | 1 | 0% | 6,203 | 3,012 | -51% | 0 | 0 | — |
case-04 | fail→pass | 10,522 | 2,323 | -78% | 1 | 1 | 0% | 1,811 | 1,455 | -20% | 0 | 0 | — |
case-05 | pass→pass | 15,730 | 9,449 | -40% | 1 | 1 | 0% | 2,387 | 2,656 | +11% | 0 | 0 | — |
case-06 | fail→fail | 8,948 | 3,156 | -65% | 1 | 1 | 0% | 1,370 | 1,625 | +19% | 0 | 0 | — |
case-07 | fail→fail | 22,313 | 11,975 | -46% | 1 | 1 | 0% | 4,826 | 3,027 | -37% | 0 | 0 | — |
case-08 | pass→pass | 14,046 | 6,024 | -57% | 1 | 1 | 0% | 2,084 | 2,009 | -4% | 0 | 0 | — |
case-09 | fail→fail | 5,251 | 2,676 | -49% | 1 | 1 | 0% | 836 | 1,511 | +81% | 0 | 0 | — |
case-10 | fail→pass | 20,755 | 10,540 | -49% | 1 | 1 | 0% | 4,285 | 2,900 | -32% | 0 | 0 | — |
case-11 | fail→pass | 7,190 | 1,907 | -73% | 1 | 1 | 0% | 1,152 | 1,381 | +20% | 0 | 0 | — |
case-12 | fail→pass | 10,636 | 2,517 | -76% | 1 | 1 | 0% | 1,663 | 1,513 | -9% | 0 | 0 | — |
case-13 | fail→pass | 6,266 | 1,810 | -71% | 1 | 1 | 0% | 882 | 1,371 | +55% | 0 | 0 | — |
case-14 | fail→pass | 9,286 | 1,724 | -81% | 1 | 1 | 0% | 1,379 | 1,353 | -2% | 0 | 0 | — |
case-15 | fail→pass | 10,990 | 2,700 | -75% | 1 | 1 | 0% | 1,619 | 1,544 | -5% | 0 | 0 | — |
case-16 | fail→pass | 10,266 | 3,422 | -67% | 1 | 1 | 0% | 1,633 | 1,651 | +1% | 0 | 0 | — |
case-17 | pass→pass | 12,777 | 2,653 | -79% | 1 | 1 | 0% | 1,919 | 1,479 | -23% | 0 | 0 | — |
case-18 | pass→pass | 14,030 | 2,430 | -83% | 1 | 1 | 0% | 2,223 | 1,454 | -35% | 0 | 0 | — |
case-19 | fail→fail | 9,633 | 5,167 | -46% | 1 | 1 | 0% | 1,475 | 2,018 | +37% | 0 | 0 | — |
case-20 | fail→pass | 12,558 | 3,959 | -68% | 1 | 1 | 0% | 1,915 | 1,719 | -10% | 0 | 0 | — |
case-21 | pass→pass | 7,726 | 1,561 | -80% | 1 | 1 | 0% | 1,187 | 1,312 | +11% | 0 | 0 | — |
case-22 | fail→fail | 9,660 | 1,944 | -80% | 1 | 1 | 0% | 1,427 | 1,413 | -1% | 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 21 counted toward the lift figure. The other 1 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 +41 percentage points is the difference between those two pass rates over the 21 comparable cases.
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.