What problem does it solve? Production incidents arrive as loose Telegram messages without structure or ownership, making it hard to move from a raw report to a verified fix. This Skill provides a disciplined workflow for evidence collection, root-cause analysis, and regression-tested fixes. ## Core Features & Use Cases - Evidence Collection: Reads RAISE bug fields, runs the incident_collect tool against a call/campaign uuid, and searches the code graph by symbol before forming any theory. - Root-Cause Documentation: Writes a root-cause document to openspec/changes/<change-id>/intake.md covering timeline, spec deviation, root cause with file:line, and blast radius, then classifies the issue as code bug, client misuse, or infra. - Regression-First Fixes: Requires a test-author agent to write a failing regression test from the incident Scenario before any fix, and verifies the fix by re-running the incident scenario with before/after evidence in the ticket and PR. - Use Case: A teammate pastes "не было end_call со стороны агента, посмотри почему" with a call uuid. The Skill walks you through collecting logs and Redis state, documenting the root cause, and shipping a minimal bugfix branch with a reproducing test. ## Quick Start Paste the incident report or call uuid and ask to investigate this production incident following the incident flow.
