Skip to content

CLAUDE.md

Latest commit

 

History

History
73 lines (51 loc) · 3.55 KB

CLAUDE.md

File metadata and controls

73 lines (51 loc) · 3.55 KB
Raw

AI Experience Framework — Root Governance File

This file is the entry point for all AI-assisted work in this project system. Read this file first. Then load files in the order specified below.

What this system is

This is a structured behavioral framework for designing and prototyping AI-enabled experiences. It separates behavioral logic (what the system should do) from experience expression (how it appears in the UI).

Your job is not only to generate screens. Your job is to reason from user intent and signals, apply the decision model, respect the control model, and produce outputs that reflect the behavioral system — not just UI compositions.

File load order

Load files in this exact sequence. Later files extend or override earlier ones where conflicts exist. Project files always take precedence over core files on project-specific decisions. Core files always take precedence on principles, constraints, and Level 0 gate rules.

Step 1 — Core framework (always load)

  1. /core/00-framework-overview.md — mental model, behavioral vs. expression layers
  2. /core/01-principles.md — non-negotiable rules, Level 0 gate, ethical constraints
  3. /core/02-pattern-library.md — reusable AI interaction patterns with usage rules
  4. /core/03-design-system-rules.md — component vocabulary, naming, design system guidance
  5. /core/04-prompting-guidelines.md — how to interpret requests, flag ambiguity, structure outputs
  6. /core/05-components.md — component behavior, states, visual rules, and code annotation standard

Step 2 — Project files (load if a project folder exists)

Check for a /projects/[project-name]/ folder. If present, load in this order:

  1. 10-project-context.md — product area, user problem, scope
  2. 11-user-jobs-and-intents.md — jobs to be done, intent states
  3. 12-signals.md — explicit and implicit signals, strength ratings
  4. 13-confidence-model.md — how confidence is defined for this product
  5. 14-decision-model.md — behavior rules, condition logic, level selection
  6. 15-control-model.md — confirm, edit, override, undo, transparency rules
  7. 16-pattern-selection.md — which patterns apply and why
  8. 17-ui-composition.md — layout, component placement, content hierarchy
  9. 18-prototype-scenarios.md — representative user states to prototype
  10. 19-output-request.md — what to generate and how to structure the response

Override rules

  • Project 13-confidence-model.md may redefine confidence thresholds for this product
  • Project 14-decision-model.md may extend or narrow the core Condition Logic A/B/C/D model
  • Project 15-control-model.md may add stricter control rules but may never remove core Level 0 constraints
  • Project 19-output-request.md extends core 04-prompting-guidelines.md — project rules layer on top, they do not replace base output structure rules

If a project file is silent on a topic, defer to the core file. If a project file contradicts a core principle or Level 0 rule, the core file wins.

Before generating any output

Run this check in order:

  1. Have I loaded all relevant core files?
  2. Have I loaded the project files if a project folder exists?
  3. Does this task require a Level 0 gate check? (See /core/01-principles.md)
  4. Do I have enough signal and intent information to select a Condition Logic?
  5. Does the output request specify how to structure my response? (See 19-output-request.md)

If any of these are unresolved, flag the gap before proceeding. Do not invent missing logic. Do not assume a confidence level without evidence.