Omnigento

Omnigento

Omnigento is a setup kit for keeping AI coding-agent instructions consistent across tools.

For a new project, Omnigento helps an AI agent create a clean instruction system from the start.

For an existing project, Omnigento helps an AI agent migrate scattered Cursor, Claude, Copilot, Codex, `AGENTS.md`, or older agent-tool instructions into one canonical instruction system.

In both cases, it keeps every provider view in sync after setup.

The user workflow is intentionally simple:

Read and follow omnigento/docs/setup.md.

You give that instruction to a filesystem-capable coding agent. The agent reads the project, runs the Omnigento scripts, creates or migrates the canonical instruction files, regenerates the provider-specific files, and verifies consistency.

You review the diff. You do not manually copy rules between five different AI tools.

flowchart TD
  A["New or existing project"] --> B["Ask an AI agent to follow omnigento/docs/setup.md"]
  B --> C{"Project state"}
  C -->|"No useful instructions yet"| D["Bootstrap clean canonical instructions"]
  C -->|"Scattered provider files exist"| E["Migrate useful rules into canonical instructions"]
  C -->|"Generated files drifted"| F["Repair generated provider views"]
  D --> G["Review one diff"]
  E --> G
  F --> G

What Problem It Solves

AI coding tools all want their own instruction format:

• GitHub Copilot uses .github/copilot-instructions.md
• Cursor uses .cursor/rules/*.mdc
• Claude Code uses CLAUDE.md, .claude/rules/*.md, and .claude/skills/*/SKILL.md
• Codex and other tools read AGENTS.md
• Older setups may have .cursorrules, GEMINI.md, .windsurfrules, .clinerules, .roo/rules/**, or .cody/**

New projects need clear rules before agents start improvising. Existing projects often already have rules scattered across provider-specific files.

Without a system, the same project rule gets copied everywhere. Then one copy changes, another does no