What problem does it solve? When a full-stack app already deployed on Rainbond fails to build, start, or serve requests, this Skill performs a bounded troubleshooting phase: it classifies the runtime state, finds the dominant blocker from events, build logs, and pod diagnostics, applies only the smallest low-risk platform-side repair, and verifies the result with fresh evidence. ## Core Features & Use Cases - Runtime State Classification: Maps live evidence to canonical states such as topology_building, runtime_unhealthy, runtime_healthy, capacity_blocked, or code_or_build_handoff_needed. - Evidence-Anchored Diagnosis: Reads component summaries, pod details, events, and build logs keyed by event_id to pinpoint blockers like missing dependencies, wrong connection values, env naming incompatibility, or ConfigMap mount failures. - Bounded Low-Risk Repair: Adds dependency edges, fixes provider connection envs, adjusts build envs, or restarts components, with a strict one-retry budget per blocker and explicit stop boundaries. - Use Case: A user says "why did the backend build fail" for an existing Rainbond app; the Skill reads component events and build logs, classifies the failure as source build failed, and hands off to code/build work instead of looping on platform edits. ## Quick Start Ask the agent to troubleshoot why the backend component of your existing Rainbond app failed to build, checking events and build logs first.
