Skip to content

PROMPT.md

Latest commit

 

History

History
114 lines (78 loc) · 3.27 KB

PROMPT.md

File metadata and controls

114 lines (78 loc) · 3.27 KB

FireLater - IT Service Management Platform

IMPORTANT: ACTION REQUIRED

You are running in autonomous mode. DO NOT just analyze or summarize. You MUST actively implement tasks from @fix_plan.md.

Your job is to:

  1. Read @fix_plan.md to find the next uncompleted task (marked with - [ ])
  2. Implement that task by writing/editing actual code files
  3. After completing a task, update @fix_plan.md to mark it done (change - [ ] to - [x])
  4. Continue to the next task

DO NOT output summaries, status reports, or analysis. WRITE CODE.

Project Overview

FireLater is a modern IT Service Management (ITSM) platform with multi-tenant architecture, built with:

  • Backend: Node.js + Fastify + PostgreSQL + Redis + BullMQ
  • Frontend: Next.js 15 + Tailwind CSS + Zustand

Current State

The platform is 94% complete with 29 of 31 phases done. All core functionality is implemented.

Primary Goal

Complete the remaining features from @fix_plan.md. Pick the first uncompleted task and implement it NOW.

Key Directories

  • /backend - Fastify API server
  • /frontend - Next.js application
  • /docker-compose.yml - Production setup
  • /docker-compose.dev.yml - Development infrastructure
  • /specs - Project requirements and specifications

Development Commands

# Start infrastructure
docker compose -f docker-compose.dev.yml up -d

# Start backend
cd backend && npm run dev

# Start frontend
cd frontend && npm run dev

# Run tests
cd backend && npm test

API Endpoints

Backend runs on http://localhost:3001 Frontend runs on http://localhost:3000 API documentation: http://localhost:3001/documentation

Constraints

  • Follow existing code patterns and naming conventions
  • Do not use emojis in code or UI
  • Prefer editing existing files over creating new ones
  • Use TypeScript strict mode
  • Follow existing Tailwind CSS class patterns

Ralph Autonomous Development

This project is configured for Ralph autonomous development loop.

Ralph Control Files

  • PROMPT.md - This file, main instructions
  • @fix_plan.md - Prioritized task list (18 tasks)
  • @AGENT.md - Build and run instructions
  • specs/requirements.md - Detailed requirements
  • .claude/settings.local.json - Pre-approved tool permissions

How Ralph Achieves Autonomy

Ralph does NOT use --dangerously-skip-permissions. Instead it uses:

  1. Pre-approved permissions in .claude/settings.local.json - specific commands are whitelisted so Claude won't prompt for approval
  2. Stdin piping - PROMPT.md is piped to claude, running in batch mode
  3. Loop orchestration - bash script handles repeated invocations with rate limiting

Running Ralph

# Start autonomous development with monitoring
ralph --monitor

# Or with custom settings
ralph --monitor --calls 50 --timeout 60

# Check status
ralph --status

Alternative: Direct Claude Code

For fully autonomous execution, you can also use:

# Option 1: Use permission bypass (use with caution in sandboxed environments)
claude --dangerously-skip-permissions

# Option 2: Specify allowed tools explicitly
claude --allowedTools "Bash(npm:*)" "Bash(git:*)" "Read" "Write" "Edit" "Grep" "Glob"

The .claude/settings.local.json in this project already pre-approves common development commands for autonomous operation.