4.1KB

Неформатований Звинувачення Історія

AGENTS.md

Purpose

Describe how AI coding agents should work in this repository.

This file is intentionally small. It acts as the repository-level router and startup guide. Detailed rules live in focused files under ./.ai/skills/ and should be loaded only when relevant to the current task.

Startup Instructions

Before working on any task:

Read this AGENTS.md.
Read the main skill index:
```
./.ai/SKILLS.md
```
Load only the relevant skill files for the current task.
Follow project-specific instructions before general best practices.
Inspect existing code before introducing new patterns.
State important assumptions before making major changes.
Prefer small, focused, reviewable changes.

Instruction Priority

When instructions conflict, follow this order:

User's current request
Safety, security, privacy, and data-loss prevention
This AGENTS.md
./.ai/SKILLS.md
Referenced files in ./.ai/skills/
Existing project conventions
General best practices

Project Summary

This project is a small PHP MVC framework called MindVisionCode PHP.

It is intentionally inspired by a Classic ASP MVC framework style:

Central dispatcher
Controllers and actions
ViewModels
Repository classes
Simple validation
Database migrations
Small, readable files
Minimal dependencies

Do not turn this project into Laravel, Symfony, Slim, CakePHP, or another large framework.

The goal is to keep the framework understandable, practical, and easy to extend.

Main Route

Always start here:

./.ai/SKILLS.md

Then load only the skill files needed for the task.

Common Skill Routes

PHP language, style, Composer, OOP:
./.ai/skills/php/SKILL.md

MVC framework architecture, routing, controllers, views, ViewModels:
./.ai/skills/mvc/SKILL.md

PDO, repositories, migrations, SQL, database safety:
./.ai/skills/database/SKILL.md

Input validation, escaping, passwords, sessions, secrets, web security:
./.ai/skills/security/SKILL.md

Tests, static analysis, quality gates, composer scripts:
./.ai/skills/testing/SKILL.md

Agent behavior, PR checklist, legacy policy, response format:
./.ai/skills/workflow/SKILL.md

Response Format

For non-trivial tasks, respond using this structure:

Goal:
Route:
Assumptions:
Plan:
Implementation:
Tests:
Risks:

For simple questions, answer directly.

Framework Change Policy

The framework core may be modified to add functionality or optimize existing code, but never silently. Any time an agent identifies a change to framework-level code (dispatcher, routing, base controller, base repository, migration runner, validation engine, autoloader, or any file under framework/), it must stop and present the following proposal to the user before writing a single line:

FRAMEWORK CHANGE PROPOSAL
==========================
Issue:
  What problem or limitation was found, and where in the framework it lives.

Proposed Change:
  What would be added, modified, or removed.

Why It Is Needed:
  The specific reason application code cannot solve this without a framework change.

Risks / Dangers:
  - Breaking changes to existing controllers, repositories, or views
  - Behavioral differences across PHP versions
  - Security surface changes
  - Performance regressions
  - Any other relevant concerns

Benefits:
  - What improves or is unlocked by the change

Alternatives Considered:
  Any application-level workarounds that were ruled out and why.

Awaiting your approval before proceeding. Reply YES to apply, NO to skip, or ask questions.

Rules:

Do not apply the change until the user explicitly approves.
If the user says NO, document the limitation as a comment or note and continue with the best available application-level workaround.
Keep framework changes small and focused — one concern per change.
After approval, note the change in the commit message so the history is clear.

Skill Feedback Rule

If project guidance is missing or unclear, suggest an update.

Suggested SKILLS.md update:
- Add/update: ...
- Reason: ...

4.1KB Неформатований Звинувачення Історія