How to use
Fill in every [PLACEHOLDER] with your specifics before pasting. The more detail you provide, the better the output.
→Fill in every [PLACEHOLDER] — vague inputs give generic outputs
→Works best in Claude Projects with the Full-Stack Architect system prompt active
→If the model stops mid-build, say "Continue from where you left off"
→For Replit: paste in the Agent chat, not the editor
Best platforms
claude.aiChatGPTReplit
The megaprompt
Build a complete AI automation tool that handles a multi-step workflow automatically.
## What This Tool Does
[DESCRIBE IN DETAIL: input → what it does → output]
Example: "Takes a list of company names, researches each one online, summarises their product, finds their pricing page, and outputs a competitive analysis spreadsheet"
## Trigger (how it starts)
- [ ] Manual trigger (user clicks a button or submits a form)
- [ ] Scheduled (runs every X hours/days)
- [ ] Webhook (triggered by another system)
- [ ] File upload (processes when user uploads a file)
- [ ] API call (triggered programmatically)
## Workflow Steps
Step 1: [What happens — input: ___, output: ___]
Step 2: [What happens — input: ___, output: ___]
Step 3: [What happens — input: ___, output: ___]
... (add as many steps as needed)
Final output: [What the user gets at the end]
## AI Actions Required
[List each step that needs AI — what model, what prompt, expected output format]
## External Integrations
- APIs to call: [LIST each API with its purpose]
- Services to connect: [Google Sheets / Notion / Slack / Email / etc]
- Files to read: [PDF / CSV / JSON / etc]
- Files to write: [CSV / JSON / PDF report / etc]
## Tech Stack
- Python (best for automation) OR Node.js (if web UI needed)
- AI: Anthropic SDK (Claude) or OpenAI SDK
- HTTP requests: httpx (Python) or axios (Node)
- Web scraping: Playwright (if needed)
- Data: pandas (Python) or simple JSON
- Queue: Simple in-memory OR Redis if scale matters
- Progress: Websocket updates to frontend OR terminal output
## User Interface
- [ ] Terminal only (scripts)
- [ ] Simple web UI with progress display
- [ ] Full dashboard with history and settings
## Output Format
The tool produces: [DESCRIBE: spreadsheet / JSON file / Slack message / email / database record / etc]
## Error Handling
- Retry failed steps up to 3 times with exponential backoff
- Log all errors to a file
- Send alert notification if entire pipeline fails
- Partial results should be saved even if pipeline fails midway
## Rate Limiting
- API calls should be throttled to avoid rate limits: [X requests per minute]
- Add delay between batch operations: [X seconds]