Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Route full software-development architecture work from product intent through design, implementation, testing, release, and operations. Use when the user asks for a complete development architecture, wants to know which Spellbook skills to combine, needs an execution path across PRD/spec/API/data/security/performance/release/SRE, or asks to turn an idea or repo into a production-ready engineering plan.
.claude/skills/majiayu000-dev-architecture-playbook/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -18% | 0% |
Use this as the lifecycle router before starting broad product or architecture work. It chooses the minimum useful skill chain, defines gates, and prevents starting implementation before the required contracts exist.
Classify the request first:
| Request | Primary Skills | Output | |---|---|---| | Idea, product direction, or market/user problem | product-discovery, prd-master | Product brief, user stories, success metrics | | Architecture or module boundaries | architecture-foundation, technical-spec, elegant-architecture | Architecture spec, boundaries, rejected alternatives | | API, auth, data, or schema contract | api-design, auth-security, database-patterns, data-contract-migrations | Versioned contracts and migration plan | | UI/product surface | frontend-design, ui-ux-pro-max, ui-design-system, playwright-automation | UX flow, component plan, visual checks | | Implementation workflow | flowguard, threads, systematic-debugging, comprehensive-testing | Bounded execution, ownership, root-cause debugging, and verification | | Quality and regression risk | comprehensive-testing, codebase-audit, vibeguard, project-health-auditor | Test matrix and risk list | | Release and operations | release-engineering, config-secrets-environments, performance-capacity, incident-slo-runbook, observability-sre, devops-excellence | Rollout, config, capacity, SLO, runbook |
Do not treat the architecture as complete until these gates are explicit:
If a gate is irrelevant, state why. Do not silently skip data, security, or rollback gates for production systems.
Use this sequence for greenfield or major refactors:
For existing repos, start with repo-agent-context-audit or codebase-audit before proposing new structure.
Return a compact plan:
textgoal: context: selected_skill_chain: architecture_gates: implementation_steps: verification_commands: release_and_ops_gates: open_risks:
Prefer the smallest chain that covers the risk. Too many skills at once usually means the scope needs to be split.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,290 | 20,204 | -5% | 1 | 1 | 0% | 3,242 | 3,510 | +8% | 0 | 0 | — |
case-02 | fail→fail | 34,007 | 25,571 | -25% | 1 | 1 | 0% | 6,257 | 4,956 | -21% | 0 | 0 | — |
case-03 | fail→pass | 15,099 | 15,794 | +5% | 1 | 1 | 0% | 2,371 | 3,321 | +40% | 0 | 0 | — |
case-04 | pass→fail | 12,529 | 12,607 | +1% | 1 | 1 | 0% | 2,394 | 3,222 | +35% | 0 | 0 | — |
case-05 | pass→fail | 6,728 | 7,453 | +11% | 1 | 1 | 0% | 1,314 | 1,987 | +51% | 0 | 0 | — |
case-06 | pass→fail | 27,028 | 9,592 | -65% | 1 | 1 | 0% | 1,327 | 2,453 | +85% | 0 | 0 | — |
case-07 | pass→pass | 12,399 | 7,871 | -37% | 1 | 1 | 0% | 1,775 | 1,874 | +6% | 0 | 0 | — |
case-08 | fail→pass | 19,271 | 12,844 | -33% | 1 | 1 | 0% | 2,940 | 2,640 | -10% | 0 | 0 | — |
case-09 | fail→pass | 30,118 | 18,555 | -38% | 1 | 1 | 0% | 4,583 | 3,646 | -20% | 0 | 0 | — |
case-10 | fail→pass | 21,767 | 14,123 | -35% | 1 | 1 | 0% | 3,631 | 2,993 | -18% | 0 | 0 | — |
case-11 | fail→pass | 19,424 | 11,073 | -43% | 1 | 1 | 0% | 2,943 | 2,406 | -18% | 0 | 0 | — |
case-12 | fail→pass | 14,712 | 12,668 | -14% | 1 | 1 | 0% | 2,233 | 2,656 | +19% | 0 | 0 | — |
case-13 | fail→pass | 21,422 | 9,704 | -55% | 1 | 1 | 0% | 3,327 | 2,219 | -33% | 0 | 0 | — |
case-14 | fail→pass | 24,880 | 14,223 | -43% | 1 | 1 | 0% | 4,142 | 2,960 | -29% | 0 | 0 | — |
case-15 | pass→pass | 12,769 | 9,414 | -26% | 1 | 1 | 0% | 1,904 | 2,233 | +17% | 0 | 0 | — |
case-16 | pass→pass | 13,844 | 6,397 | -54% | 1 | 1 | 0% | 2,007 | 1,849 | -8% | 0 | 0 | — |
case-17 | pass→pass | 9,705 | 5,365 | -45% | 1 | 1 | 0% | 1,441 | 1,548 | +7% | 0 | 0 | — |
case-18 | fail→pass | 12,264 | 5,029 | -59% | 1 | 1 | 0% | 1,793 | 1,457 | -19% | 0 | 0 | — |
case-19 | fail→pass | 9,445 | 2,665 | -72% | 1 | 1 | 0% | 1,456 | 1,077 | -26% | 0 | 0 | — |
case-20 | fail→pass | 22,299 | 13,122 | -41% | 1 | 1 | 0% | 3,554 | 2,656 | -25% | 0 | 0 | — |
case-21 | fail→pass | 21,402 | 13,531 | -37% | 1 | 1 | 0% | 3,247 | 2,773 | -15% | 0 | 0 | — |
case-22 | fail→pass | 22,409 | 13,074 | -42% | 1 | 1 | 0% | 3,537 | 2,803 | -21% | 0 | 0 | — |
case-23 | fail→pass | 14,570 | 14,227 | -2% | 1 | 1 | 0% | 2,257 | 2,893 | +28% | 0 | 0 | — |
case-24 | fail→pass | 22,535 | 15,156 | -33% | 1 | 1 | 0% | 3,572 | 3,044 | -15% | 0 | 0 | — |
case-25 | fail→pass | 28,135 | 13,006 | -54% | 1 | 1 | 0% | 4,220 | 2,682 | -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. 25 cases were attempted. The headline lift of +56 percentage points is the difference between those two pass rates over the 25 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.