Based in @DaC and @Fachten help, I have made my set of prompts below to use at prototyper. The step 5 and the “context clear” parts unfortunately are not possible at prototyper screen.
I pasted item by item at the prototyper and it worked.
We will reset your memory between sessions, so we need a way to retain project context and progress across our interactions.
For that, create a `context/` directory in the root of the project.
Please follow the steps below carefully.
## 1. 📚 Create a file called "PROJECT_CONTEXT_HISTORY.md"
In this file, create a structured summary of the entire project based on our conversations and codebase. Include:
- **Project Objectives and Vision**
- **Technical Stack and Architecture**
- **Features Implemented**
- **Features Currently Under Development**
- **Planned Features / Future Work**
- **Design Decisions and Constraints**
- **Open Questions / Assumptions**
- **List the complete data structure of the app**
- **List all firebase rules, cloud functions and APIs**
- **Other Important Observations**
Save this to:
`context/PROJECT_CONTEXT_HISTORY.md`
## 2. 📊 Create a file called "PROJECT_PROGRESS.md"
Track project status using a clear, simple format:
DONE:
[Completed task 1]
[Completed task 2]
WORKING:
[In-progress task 1]
[In-progress task 2]
NEXT:
[Upcoming task 1]
[Upcoming task 2]
📄 Save this to:
`context/PROJECT_PROGRESS.md`
3. 🧭 Create a file called "PROJECT_INSTRUCTIONS.md"
Create internal instructions for yourself (the AI) with clear guidance on how to manage project context in subsequent sessions. These instructions **must** include:
### 🔁 When to Update
- After completing a major task.
- After deciding something new about the project.
- After a major discussion or change in scope.
### 📄 How to Update
- Update `PROJECT_CONTEXT_HISTORY.md` with new relevant info.
- Add or move items in `PROJECT_PROGRESS.md` according to task status.
- Keep each file tidy and avoid duplicated or outdated info.
### 📌 Startup Routine
Before responding in any future session, always:
1. Load and parse the content of `context/PROJECT_CONTEXT_HISTORY.md`.
2. Load and review `context/PROJECT_PROGRESS.md`.
3. Use this context to inform responses, code suggestions, and planning.
4. If context files are missing or outdated, ask the user to clarify or update them.
### ⚠️ Important Commands (to be followed by the AI)
- `ALWAYS read context/*.md files before starting any session.`
- `NEVER assume project status without checking PROJECT_PROGRESS.md.`
- `Update context files whenever a significant change happens.`
- `Log changes made by you (the AI) in PROJECT_CONTEXT_HISTORY.md.`
- `If user gives you a new context, integrate and persist it immediately.`
📄 Save this to:
`context/PROJECT_INSTRUCTIONS.md`
## 4. 🧠 Create a file called "1st.md" (Session Boot Prompt)
Generate a prompt that can be used to kick off any future session with full project awareness. It should tell the AI to:
You are now rejoining a project with prior context.
Read the files in the context/ directory:
- PROJECT_CONTEXT_HISTORY.md
- PROJECT_PROGRESS.md
- PROJECT_INSTRUCTIONS.md
Load all relevant project information into memory.
Use the loaded context to inform all answers, decisions, suggestions, and code actions.
Ask for clarification if anything seems missing or outdated.
Follow the update routines specified in PROJECT_INSTRUCTIONS.md.
📄 Save this to:
`context/1st.md`
5. MANUAL STEP IN THE TERMINAL!!!
-
Create the folder “context_backup” inside “context” like “context/context_backup”
-
use the script below in the terminal:
timestamp=$(date +%Y%m%d%H%M%S); mkdir -p "/home/user/studio/context/context_backup/" && cp "/home/user/.idx/ai/capra-context-state.json" "/home/user/studio/context/context_backup/capra-context-state-${timestamp}.json" && cp "/home/user/.idx/ai/capra-thread.json" "/home/user/studio/context/context_backup/capra-thread-${timestamp}.json"
## 6. Add the backup folder to .gitignore:
"# prototyper
/context/context_backup"
IMPORTANT: This prevents the context_backup folder and the backup files within it from being used as input for the App prototyper agent context. If you don’t follow this step, you’ll hit a snag “The input token count exceeds the maximum number of tokens allowed.”