For Coding Agents

Execution contract

Agents should read docs URLs as canonical, avoid destructive operations by default, and confirm app context before changes.

• Read `llms.txt` first, then `llms-full.txt`
• Use `/openapi.json` for endpoint details
• Run `archastro status` before any write command
• Prefer read commands when context is incomplete

Claude Code + ArchAstro CLI scenario

This sequence is designed for autonomous coding agents operating in a repo terminal.

1. Verify auth and app context.
2. Read existing resources to avoid duplication.
3. Perform exactly one scoped change.
4. Re-read resource and report machine-readable diff summary.

archastro status
archastro get agents
archastro get agent <agent-id>
# apply one scoped change command here
archastro get agent <agent-id>

{
  "action": "update-agent",
  "target": "agent_<id>",
  "status": "success",
  "verification": [
    "archastro get agent <agent-id>"
  ]
}