Runtime Adapters

SkillAI & models

Lets one set of agent instructions work across Codex, Claude Code, Gemini CLI, Cursor, and AGENTS.md by creating an adapter for each. Use it when a generated repo needs to support several AI tools, so you keep a single source of instructions instead of separate copies for every tool.

Available today. Use it from your connected AI after setup.

After adding it, tell your AI which tools you want to support and it will create the matching adapters from your agent's core instructions.

Then ask your AI: use the Runtime Adapters skill

What your AI can do with it

  • Create adapters for Codex, Claude Code, Gemini CLI, Cursor, and AGENTS.md
  • Build every adapter from a single source of agent instructions
  • Set up a generated repo to work with several AI tools
  • Keep instructions in one place instead of copying them for each tool

What this skill tells your AI

The instructions your AI receives, as published by agentlas-ai/agentlas-os in skills/runtime-adapters/SKILL.md and read by ahel’s review.

Rules

  • AGENTS.md is canonical.
  • .claude/, .codex/, .gemini/, .cursor/, and root skills/ are thin adapters or mirrors.
  • .agentlas/global-commands.json records the command each adapter exposes.
  • Adapter text should point back to AGENTS.md.
  • Do not claim identical behavior across runtimes. Say the canonical core is portable and each adapter maps it into local conventions.

Required Adapters

  • Codex: plugin manifest plus skill and commands/<slug>.md.
  • Claude Code: command, agent, and skill adapter under .claude/commands/.
  • Gemini CLI: GEMINI.md plus .gemini/commands/<slug>.toml.
  • Generic: root AGENTS.md with the command alias documented.

Signals

GitHub stars
1k
Forks
103
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
runtime-adapters
Source
github.com/agentlas-ai/agentlas-os