Diagnose an error and recover
Use the rules in SKILL.md and core.md and establish the live target/deployed release. Read its README's "Troubleshooting" and "Before you resume". Use knowledge.md for factual EVS questions while keeping the deployment context.
Read logs, stack events, and environment/host state to locate the failure. Report the observed stage and supporting evidence, resources still active or billable, the needed correction, and the proposed recovery's exact target, commands, scope, and destructive effects.
If EDO cannot perform the correction through its documented capabilities, explain what must resolve it. Do not replace or patch the tool.
Once recovery is agreed, re-check deployed state, perform the correction and documented resume, and record results. An ambiguous result remains unresolved until live evidence clarifies it. Return to monitor.md.