npx agentspec-cli install 00e33978-bb8e-4644-a48d-0732a660007c --file-type agents-md→ .agents/
# AGENTS.md ## Project OpenReef is an offline-first AI agent mobile application built with Flutter. Core system includes: - Agent engine (planning + execution) - Memory system (context + persistence) - Tool/skill execution layer - MCP integration - Flutter UI layer The app is NOT a generic Flutter app. It is an agent-driven system. All changes must respect this. --- ## Architecture (STRICT) Follow docs/02_system/system-overview.md exactly. Allowed top-level structure: - lib/agent/ - lib/context/ - lib/mcp/ - lib/skills/ - lib/memory/ - lib/tools/ - lib/ui/ Rules: - Do NOT create new top-level folders - Do NOT move modules between domains - Do NOT redesign architecture - Do NOT introduce alternative patterns If something does not fit → adapt inside existing modules --- ## Code Rules (STRICT) - Do NOT modify unrelated files - Do NOT perform large refactors - Do NOT add new dependencies unless absolutely required - Prefer extending existing classes over creating new systems - Keep functions small and focused - Avoid over-engineering Naming: - files → snake_case - classes → UpperCamelCase - methods/variables → lowerCamelCase --- ## Agent System Rules (CRITICAL) - Agent logic MUST stay inside `lib/agent/` - Memory logic MUST stay inside `lib/memory/` - Tool execution MUST stay inside `lib/tools/` - UI MUST NOT contain business logic Do NOT mix these layers. --- ## Development Commands: flutter pub get flutter analyze flutter test flutter run Before finishing any task: - run analyzer - ensure no warnings - ensure code compiles --- ## Testing - Place tests under `test/` mirroring `lib/` - Prioritize: - agent planning logic - tool routing - memory writes --- ## Verification After every implementation: 1. Run static analysis: flutter analyze 2. Run tests: flutter test 3. If there are errors or warnings: - Fix them immediately - Do NOT leave broken code 4. Ensure: - Code compiles - No analyzer warnings - No failing tests 5. Do NOT proceed to next task until all checks pass --- ## What NOT to do - Do NOT invent new architecture - Do NOT add random abstractions - Do NOT introduce unnecessary state management solutions - Do NOT rewrite working code - Do NOT guess behavior — follow docs/ --- ## Source of Truth Primary: - docs/02_system/system-overview.md Secondary: - other docs/ files If conflict exists → architecture doc wins PRs should include a clear summary, linked issue or spec section, testing notes, and screenshots or recordings for UI changes. If a change updates architecture or behavior, also update the relevant file in [`docs/`](/D:/Software/Flutter/OpenReef/docs).
AGENTS.md
This document provides guidance for any AI agent working with code in this repository. Agents are expected to contribute to the project effectively and safely, based on the information provided below.
by AgentHub Crawler
AGENTS.md
The repository is a Flutter application that follows the Flutter app architecture guidance with explicit layering, MVVM-style presentation, dependency injection via `get_it`, and modular agent resources under `.agents/`.
by AgentHub Crawler
Project Overview
Modern application with Flutter frontend and Rust backend.
by AgentHub Crawler
Agents.md – Guidance for OpenAI Codex
Sign in to leave a comment.
This `Agents.md` file provides detailed instructions for OpenAI Codex and other AI agents collaborating on this repository. It includes project structure, conventions, testing, and update policies.
by AgentHub Crawler