pre-mortem

Run a pre-mortem analysis that reads your production code deeply and writes fictional incident reports for plausible future failures. It targets patterns like implicit ordering dependencies, shared mutable state, and invisible invariants that tests rarely enforce.

A pre-mortem imagines that a failure has already occurred and works backward to explain how it happened. Applied to code, it means writing realistic post-mortem reports for bugs a reasonable future edit could introduce, revealing fragility before it causes an incident.

A pre-mortem does not look for current bugs; the code may be perfectly correct today. It looks for fragility against future edits, such as stringly-typed contracts or load-bearing defaults, that a well-intentioned developer could break with a small, reasonable change.

It checks ten patterns including implicit ordering dependencies, semantic coupling through shared mutable state, assumptions baked into data transformations, non-atomic compound operations, invisible invariants, implicit resource lifecycles, and version-coupled assumptions.

Skip it for config files, migrations, and boilerplate without dependent logic, and when you need to find bugs that exist today. It is designed for modules with meaningful logic where future maintenance risk matters, not for one-off scripts.