AgentMD - AI Agent Rules Library
Create better AI coding workflows with curated rule templates, framework-specific presets, and reusable standards for secure, testable, maintainable code.
Author
agent_rules_bot
Agent
Gemini CLI
Language
Elixir
Framework
Phoenix
Testing rules
Security rules
Prefer minimal diffs
Architecture rules
# Gemini CLI · Elixir + Phoenix Generated February 10, 2026 Agent: Gemini CLI Language: Elixir Framework: Phoenix ## Global rules - Start by reading existing project instructions and architecture docs before changing behavior. - Prefer minimal diffs and preserve existing code style and naming conventions. - State assumptions and call out risks before making breaking or cross-module changes. - When behavior changes, add or update tests in the closest existing test layer. - Never commit secrets, tokens, credentials, or generated private keys. Start by reading existing project instructions and architecture docs before changing behavior. Prefer minimal diffs and preserve existing code style and naming conventions. State assumptions and call out risks before making breaking or cross-module changes. When behavior changes, add or update tests in the closest existing test layer. Never commit secrets, tokens, credentials, or generated private keys. ## Agent-specific - Prefer deterministic, reproducible shell commands over manual workflows. - Structure output as actions, results, and next steps. - When generating code, prioritize maintainability over cleverness. ## Language-specific - Model workflows as small pure functions and supervised processes. - Keep GenServer state minimal and explicit. - Prefer pattern matching and tagged tuples for error handling. ## Framework-specific - Use contexts to enforce domain boundaries. - Keep controllers/channels thin and move logic into context modules. ## Options ### Testing rules - Require tests for changed behavior or document why tests were not added. - Run the smallest relevant test subset first, then broaden if needed. ### Architecture rules - Preserve module boundaries and avoid leaking internal abstractions. - Prefer incremental refactors over large rewrites. ### Security rules - Validate and sanitize all user-controlled input. - Apply least-privilege defaults for data access and actions. ### Prefer minimal diffs - Limit changes to the smallest set of files and lines needed. - Defer unrelated cleanup to separate follow-up changes.