ai-code-blindspots
version: 0.1.0
ai-code-blindspots
Overview
Catches the omissions and fabrications AI is most likely to leave out / invent when writing code — the things that compile, look correct in a quick read, and silently break at runtime or in production. Complements OMO's built-in remove-ai-slops skill, which hunts bloat / over-engineering / dead flexibility (code that should not exist); this skill hunts missing checks and fabricated references (code that should exist but doesn't, OR references things that don't exist).
| Skill | Hunts | Example finding |
|---|---|---|
remove-ai-slops |
over-engineering, smells, bloat | "this abstraction has one implementation — delete it" |
ai-code-blindspots (this) |
missing checks, silent failures, deprecated calls | "this .map() is called on a possibly-undefined field — add a guard" |
The two skills do not overlap; running both back-to-back is the intended workflow. Each finds bugs the other structurally cannot.
Output: .omo/blindspots-reports/<diff-hash>/ai-blindspots-report.md(process-grade: ephemeral working artifact, not a git-tracked audit log; gitignored per diff — <diff-hash> defined in §5) — listing every finding with file:line + a one-line fix. Findings cite the checklist class (1-8 below) so the reader can audit coverage.