ai-native-reviewlisted

# AI-Native Design Review ## When to Trigger - Before designing a new Skill, tool, Web App, or feature - During implementation when semantic work is drifting into rules, regex, templates, or click-heavy workflows - After an app, Skill, agent, or tool exists and needs an AI-native audit - When asked to review an existing tool for AI-native quality - When user says "check if this is AI-native" or "review this design" ## Two Dimensions Review two independent but complementary dimensions: | Dimension | Core Question | Healthy Pattern | |-----------|---------------|-----------------| | **AI-native intelligence** | Does the model make the judgment calls that require understanding? | LLM judgment + evidence + confidence | | **AI-native interaction** | Can users express the outcome they want while the system plans and executes the workflow under appropriate human control? | Goal-oriented input + orchestration + visible progress + confirmation | A tool may pass one dimension and fail the other. Do not average them into one score. Report each dimension separately. ## Review Process ### Step 0: Quick Routing Route the review by dimension: 1. **Core function is purely deterministic?** (file management, format conversion, CRUD) → Intelligence audit may be `N/A` 2. **In a compliance/security/latency-critical domain?** (regulatory checks, real-time API) → Mark relevant deterministic rules as justified 3. **Low complexity?** (< 3 subtasks, no semantic understanding needed) → Use a