Instruction file imported from cfxdevkit/devkit-workspace-example (.github/instructions/devkit-diagnostics.instructions.md). Copyright stays with the author.

DevKit Diagnostics — Stack and Runtime Diagnosis

Use this when the runtime context is ambiguous, stack readiness is unclear, or backend/MCP behavior seems inconsistent.

Diagnosis order

1. conflux_status — read Next step field carefully
2. agent_workspace_context — confirm which workspace/runtime context the agent is in
3. local_stack_status — full stack readiness check
4. agent_runbook_execute with local_stack_doctor — when the issue is still unclear after step 3
5. backend_health — narrow backend-only health check
6. agent_operations_recent — inspect prior failed actions before retrying

Rules

• Prefer backend-first diagnosis over compose-first diagnosis.
• Treat missing Docker/compose visibility as diagnostic context, not proof that the workspace backend is unusable.
• Do not move into deploy or DEX operations until readiness is explicit from local_stack_status.
• If a tool call returns an unexpected error, check agent_operations_recent for context before retrying.

Node lifecycle recovery

When diagnosis is complete

• Confirm stack is ready with local_stack_status before proceeding to deploy or DEX operations.
• If accounts are needed, verify with conflux_accounts.