Generate divergent architectural proposals with contract-validated JSON and convergence detection.
npx skills add https://github.com/luyzkk/Anti-Vibe-Coding --skill design-twice
Please help me install this Agent Skill. Skill: design-twice Source: https://github.com/luyzkk/Anti-Vibe-Coding/tree/main/skills/design-twice Command: npx skills add https://github.com/luyzkk/Anti-Vibe-Coding --skill design-twice
Prevents premature convergence on a single architecture by forcing multiple genuinely different proposal directions early, so the decision reflects your constraints instead of early assumptions.
.claude/decisions.md with a DT- prefix.Ask design-twice to explore an architecture direction by running /anti-vibe-coding:design-twice with your problem or feature description as the argument.
High-intent search queries and answers about installing and using this skill.
To generate divergent architectures, you can use a parallel sub-agent approach that extracts codebase constraints and produces 3 to 5 distinct proposals. This forces comparative evaluation before recording a final structural decision.
Premature convergence happens when a single architecture is chosen based on early assumptions. Avoid it by generating multiple genuinely different proposal directions early, ensuring your final decision reflects actual project constraints.
You can record architectural tradeoffs by consolidating parallel proposals into a comparison table and registering the selected outcome. This creates a documented decision log with a specific prefix for future reference.
Yes, you can compare different data schema and frontend approaches. The process applies constraint-first context extraction to generate complete, isolated architectural directions for both existing and greenfield codebases.
It works with both existing and greenfield codebases. The process reads your current stack, libraries, patterns, and relevant entities to build shared constraints before generating isolated architectural proposals.
You should avoid generating parallel divergent architecture proposals for simple, deterministic changes that lack structural tradeoffs. The approach is designed for complex decisions involving technology choices, data schemas, or system structure.
peppone-choiValidate game codebase parity against architectural documentation.
crewrigResolve competing proposals into governed decision records with SPECS follow-up issues.
agntcyConverts JSON Schema files to Python objects.
rd162Automate iterative critique and revision loops with isolated CRITIC and AUTHOR agents.