AGENTS.md

Project Overview

Language: ruby Type: web Instruction Level: standard Architecture: layered

Development Guidelines

This project follows structured coding guidelines across multiple categories:

Language

  • basics
  • testing

Testing

  • unit-fundamentals
  • unit-mocking
  • basics

Security

  • injection
  • auth-jwt
  • secrets

Performance

  • basics
  • async

API Design

  • basics
  • rest
  • pagination

Code Style

  • naming
  • organization

Error Handling

  • strategy
  • basics

Architecture

  • principles
  • layers
  • patterns

DevOps

  • ci-cd
  • observability

Best Practices

  • planning
  • documentation
  • code-review

Design Patterns

  • base-patterns
  • data-access
  • domain-logic

Commands

Install dependencies:

# Check package.json for package manager
npm install
# or yarn install
# or pnpm install
# or bun install

Run tests:

# Check package.json scripts section
npm test

Build:

# Check package.json scripts section
npm run build

Code Style

See tool-specific instruction files for detailed code style guidelines:

  • Claude Code: CLAUDE.md
  • GitHub Copilot: .github/copilot-instructions.md
  • Antigravity: .agent/rules/instructions.md
  • OpenAI Codex: .codex/instructions.md

Architecture

This project follows layered architecture. See architecture guidelines in tool-specific files.

Agentic Profile

Profile level: standard

Enabled surfaces:

  • Project memory: CLAUDE.md (passive, stable)
  • Slash commands: .claude/commands/*.md (guided, stable)

Limits and gated surfaces:

  • Project subagents: full+; Use focused agents with narrow review jobs to avoid noisy or contradictory reviews.
  • Project skills: full+; Skills should use progressive disclosure and keep heavy references out of the main prompt.
  • Lifecycle hooks: full+; Hooks execute shell commands automatically. Do not generate full test-suite hooks; use /check for validation.

Side-effecting features such as hooks, MCP, plugin setup scripts, or executable tool configuration must stay local, deterministic, and explicitly reviewed before use.

Testing

Follow testing guidelines in tool-specific instruction files.

Git Workflow

  • Write clear commit messages
  • Follow conventional commits if configured
  • Run tests before pushing
  • Keep PRs focused and reviewable

Workflows

Use these slash commands for structured SDLC delivery:

Flow: /spec → /research → /plan → /build → /check → /ship

  • /spec — Capture the full specification for a feature or task before any code is written.
  • /research — Analyze the active spec with internal codebase scanning and external web research.
  • /plan — Produce a phased, checkpoint-driven implementation plan based on the spec and research findings.
  • /build — Execute the next (or a specified) phase of the current implementation plan.
  • /check — Verify the current implementation against the active spec — tests, code review, and regression check.
  • /ship — Pre-flight wrap-up — verify everything is ready, then draft a PR description.

Generated by aicgen - Universal AI agent instructions following the AGENTS.md standard