jean-philippe
3fae2fcbe1
This is the initial commit for the CODEX_ADK project, a full-stack AI agent management platform featuring: BACKEND (ASP.NET Core 8.0): - CQRS architecture with 6 commands and 7 queries - 16 API endpoints (all working and tested) - PostgreSQL database with 5 entities - AES-256 encryption for API keys - FluentValidation on all commands - Rate limiting and CORS configured - OpenAPI/Swagger documentation - Docker Compose setup (PostgreSQL + Ollama) FRONTEND (Flutter 3.x): - Dark theme with Svrnty branding - Collapsible sidebar navigation - CQRS API client with Result<T> error handling - Type-safe endpoints from OpenAPI schema - Multi-platform support (Web, iOS, Android, macOS, Linux, Windows) DOCUMENTATION: - Comprehensive API reference - Architecture documentation - Development guidelines for Claude Code - API integration guides - context-claude.md project overview Status: Backend ready (Grade A-), Frontend integration pending 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
100 lines
2.4 KiB
Markdown
100 lines
2.4 KiB
Markdown
# MANDATORY RESPONSE PROTOCOL
|
||
|
||
**Claude must strictly follow this protocol for ALL responses in this project.**
|
||
|
||
---
|
||
|
||
## 🗣️ Response Protocol — Defined Answer Types
|
||
|
||
Claude must **always** end responses with exactly one of these two structured formats:
|
||
|
||
---
|
||
|
||
### **Answer Type 1: Binary Choice**
|
||
Used for: simple confirmations, proceed/cancel actions, file operations.
|
||
|
||
**Format:**
|
||
|
||
(Y) Yes — [brief action summary]
|
||
|
||
(N) No — [brief alternative/reason]
|
||
|
||
(+) I don't understand — ask for clarification
|
||
|
||
|
||
**When user selects `(+)`:**
|
||
Claude responds:
|
||
> "What part would you like me to explain?"
|
||
Then teaches the concept step‑by‑step in plain language.
|
||
|
||
---
|
||
|
||
### **Answer Type 2: Multiple Choice**
|
||
Used for: technical decisions, feature options, configuration paths.
|
||
|
||
**Format:**
|
||
|
||
(A) Option A — [minimalist description]
|
||
|
||
(B) Option B — [minimalist description]
|
||
|
||
(C) Option C — [minimalist description]
|
||
|
||
(D) Option D — [minimalist description]
|
||
|
||
(+) I don't understand — ask for clarification
|
||
|
||
|
||
**When user selects `(+)`:**
|
||
Claude responds:
|
||
> "Which option would you like explained, or should I clarify what we're deciding here?"
|
||
Then provides context on the decision + explains each option's purpose.
|
||
|
||
---
|
||
|
||
### ⚠️ Mandatory Rules
|
||
1. **No text after the last option** — choices must be the final content.
|
||
2. Every option description ≤8 words.
|
||
3. The `(+)` option is **always present** in both formats.
|
||
4. When `(+)` is chosen, Claude shifts to teaching mode before re‑presenting options.
|
||
5. Claude must include `(always read claude.md to keep context between interactions)` before every option set.
|
||
|
||
---
|
||
|
||
### Example 1 (Binary)
|
||
|
||
We need to initialize npm in your project folder.
|
||
|
||
(always read claude.md to keep context between interactions)
|
||
|
||
(Y) Yes — run npm init -y now
|
||
|
||
(N) No — show me what this does first
|
||
|
||
(+) I don't understand — explain npm initialization
|
||
|
||
|
||
### Example 2 (Multiple Choice)
|
||
|
||
Choose your testing framework:
|
||
|
||
(always read claude.md to keep context between interactions)
|
||
|
||
(A) Jest — popular, feature-rich
|
||
|
||
(B) Vitest — faster, Vite-native
|
||
|
||
(C) Node test runner — built-in, minimal
|
||
|
||
(D) Skip tests — add later
|
||
|
||
(+) I don't understand — explain testing frameworks
|
||
|
||
|
||
---
|
||
|
||
**This protocol ensures:**
|
||
- You always have an escape hatch to learn.
|
||
- Claude never assumes your technical knowledge.
|
||
- Every interaction has clear, actionable paths.
|