What problem does it solve? When working on the Sonder Runtime codebase, engineers risk re-attempting fixes that already failed, undoing deliberate reverts, or misreading strangely-shaped guards without knowing their history. This Skill provides chronological institutional memory of every major investigation, dead end, rejected fix, and revert so you can answer "why is this code written this way" before changing it. ## Core Features & Use Cases - Incident index with verified status: Thirteen documented incidents, each with symptom, root cause, evidence commits, and a status (fixed, open, or policy) re-verified against a specific commit. - Fix discipline checklist: A distilled six-step sequence covering reproduction through production paths, mutation-proving guards, corroborating zero-counts, and re-verifying file:line anchors and git lineage. - Re-verification one-liners: Ready-to-run shell commands that confirm whether each incident's fix is still present in the current tree. - Use Case: Before re-attempting a fix that a code comment hints was tried before, consult this Skill to learn that shell_run was wholesale-reverted as a safety policy, or that the smoke gate was once a constant-true assertion, so you do not repeat the mistake. ## Quick Start Ask why a specific guard or check in the Sonder Runtime repo is shaped the way it is, or whether a particular fix has been tried before.
