Skip to main content

Use with Claude Code

The Open Agentic Platform treats the AI agent development environment as first-class infrastructure. The repository ships with native Claude Code integration located in the .claude/ directory.

First-Class Infrastructure​

The .claude/ directory is not just a collection of helpful prompts; it defines the orchestrated workflows, safety rules, and agent behaviors that build the platform itself.

  • .claude/agents/: Defines the specific personas used in the development pipeline.
  • .claude/skills/: Contains the commands (skills) available to the agents.
  • .claude/rules/: Defines the behavioral constraints enforced by the orchestrator.

The AGENTS.md Protocol​

The initialization of any agent session is governed by the AGENTS.md file at the repository root. This file acts as the cross-agent session-init protocol authority (compatible with the AAIF/Linux Foundation AGENTS.md standard).

When you start a session using the /init skill, it reads AGENTS.md to load the context, recent activity, and governed reads of the registry and codebase index.

Strict Read Discipline​

A critical rule enforced during initialization (Spec 103) is that the init protocol MUST NOT parse compiled JSON artifacts directly. All structural and lifecycle data must be retrieved through the spec-spine consumer binaries. If the index is stale, the /init protocol will instruct the developer to run the appropriate make targets rather than attempting to bypass the compiler.