Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Selects architecture paradigm via research before scaffolding. Use when architecture is undecided and the choice needs justification and documentation.
.claude/skills/athola-architecture-aware-init/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 66% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 62% | 0% |
Project initialization that combines online research, archetype selection, template customization, and decision documentation into one workflow. Use this skill when the architecture is undecided and the choice deserves justification.
Use instead of project-init when architecture is undecided. Use before project-specification to establish the architectural foundation.
arch-init:research-completed: Online research completedarch-init:paradigm-selected: Architecture paradigm chosenarch-init:templates-customized: Templates adapted to paradigmarch-init:decision-recorded: ADR createdLoad modules/research-flow.md for the full project-context questionnaire and the three-tier search strategy. Output: a synthesis brief that feeds Step 3.
Load modules/paradigm-selection.md for the decision matrix (team size by domain complexity) and the special-case overrides (streaming, serverless, microkernel, etc.). Two routes:
archetypes:architecture-paradigms skill for guidedexploration.
Load modules/scaffold-generation.md for the paradigm-specific directory layouts (Functional Core / Hexagonal / Microservices shown; others delegated to the corresponding archetypes:architecture-paradigm-{name} skill) and the ADR template.
After completing the workflow, the project has:
tooling, dependency hints)
implementation guidance
The interactive workflow above is the default. For automation, load modules/script-integration.md for the three Python helpers under plugins/attune/scripts/ (architecture researcher, template customizer, full interactive flow) and library-style import examples.
This skill enhances /attune:project-init by adding an architecture-selection phase before scaffolding:
bash# Standard initialization (no architecture decision) /attune:project-init --lang python --name my-project # Architecture-aware initialization /attune:brainstorm # Explore project needs Skill(architecture-aware-init) # Select architecture /attune:project-init --arch <paradigm> --name my-project
User: "I'm creating a Python web API for a fintech application. Team of 8 developers, complex business rules, need high security and audit trails."
engineers, security and auditability requirements.
architecture, CQRS+ES Python examples.
Event Sourcing.
module, event store, aggregate patterns, projection handlers.
rules, regulatory compliance).
Result: project initialized with paradigm-appropriate structure and clear decision rationale.
Skill(archetypes:architecture-paradigms): paradigm catalogSkill(archetypes:architecture-paradigm-*): per-paradigmimplementation guidance
Skill(attune:project-brainstorming): ideation beforearchitecture
Skill(attune:project-specification): requirements afterthe paradigm is chosen
/attune:project-init: basic project initialization/attune:blueprint: architecture planning after paradigmselection
plugins/archetypes/README.md: full paradigm referencearch-init:research-completed, arch-init:paradigm-selected, arch-init:templates-customized, arch-init:decision-recorded.
chosen, referencing evidence from the Step 2 research synthesis.
archetypes:architecture-paradigm-* skill (verified by listing the created directories).
with trade-offs and waits for explicit user selection rather than defaulting silently.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 36,050 | 24,395 | -32% | 1 | 1 | 0% | 5,845 | 5,122 | -12% | 0 | 0 | — |
case-02 | fail→fail | 31,625 | 19,237 | -39% | 1 | 1 | 0% | 6,217 | 4,351 | -30% | 0 | 0 | — |
case-03 | fail→pass | 31,368 | 29,485 | -6% | 1 | 1 | 0% | 6,209 | 6,326 | +2% | 0 | 0 | — |
case-04 | pass→pass | 12,200 | 10,776 | -12% | 1 | 1 | 0% | 2,241 | 3,218 | +44% | 0 | 0 | — |
case-05 | pass→pass | 35,928 | 21,668 | -40% | 1 | 1 | 0% | 6,188 | 4,816 | -22% | 0 | 0 | — |
case-06 | pass→pass | 16,723 | 14,062 | -16% | 1 | 1 | 0% | 2,370 | 3,113 | +31% | 0 | 0 | — |
case-07 | fail→pass | 13,932 | 18,823 | +35% | 1 | 1 | 0% | 2,121 | 3,116 | +47% | 0 | 0 | — |
case-08 | fail→pass | 9,621 | 8,033 | -17% | 1 | 1 | 0% | 1,419 | 2,351 | +66% | 0 | 0 | — |
case-09 | fail→pass | 14,010 | 15,039 | +7% | 1 | 1 | 0% | 2,345 | 3,800 | +62% | 0 | 0 | — |
case-10 | fail→pass | 5,406 | 4,208 | -22% | 1 | 1 | 0% | 823 | 1,859 | +126% | 0 | 0 | — |
case-11 | fail→pass | 9,558 | 2,458 | -74% | 1 | 1 | 0% | 1,406 | 1,571 | +12% | 0 | 0 | — |
case-12 | fail→fail | 8,198 | 1,355 | -83% | 1 | 1 | 0% | 1,184 | 1,333 | +13% | 0 | 0 | — |
case-13 | fail→pass | 9,252 | 2,519 | -73% | 1 | 1 | 0% | 1,567 | 1,548 | -1% | 0 | 0 | — |
case-14 | fail→pass | 12,091 | 1,359 | -89% | 1 | 1 | 0% | 2,184 | 1,347 | -38% | 0 | 0 | — |
case-15 | fail→pass | 11,769 | 1,928 | -84% | 1 | 1 | 0% | 1,853 | 1,447 | -22% | 0 | 0 | — |
case-16 | pass→pass | 5,991 | 3,166 | -47% | 1 | 1 | 0% | 1,015 | 1,670 | +65% | 0 | 0 | — |
case-17 | pass→pass | 18,231 | 15,109 | -17% | 1 | 1 | 0% | 2,760 | 3,519 | +28% | 0 | 0 | — |
case-18 | fail→pass | 17,814 | 17,180 | -4% | 1 | 1 | 0% | 2,729 | 3,933 | +44% | 0 | 0 | — |
case-19 | fail→pass | 6,322 | 1,895 | -70% | 1 | 1 | 0% | 852 | 1,434 | +68% | 0 | 0 | — |
case-20 | fail→pass | 7,777 | 3,132 | -60% | 1 | 1 | 0% | 1,144 | 1,642 | +44% | 0 | 0 | — |
case-21 | fail→pass | 12,095 | 4,966 | -59% | 1 | 1 | 0% | 1,836 | 1,970 | +7% | 0 | 0 | — |
case-22 | pass→pass | 9,846 | 7,130 | -28% | 1 | 1 | 0% | 1,492 | 2,378 | +59% | 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 +64 percentage points is the difference between those two pass rates over the 22 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.