ai-flow /learn: learnings from d3mlabs/dev#133

.cursor/rules/learnings-index.mdc

@@ -55,6 +55,13 @@ propose a retirement, a consolidation, or a glob-scoped sub-index split.
  render re-serializes its output and any mis-split compounds on the next
  pull. → .cursor/skills/learnings/plan-parse-normalizes/
 
+## style
+
+- [style/comment-line-width] Code comments fill the 120-char rubocop
+ width; don't pattern-match the legacy ~80-column wrapping left unswept
+ from the 80→120 expansion (sweep: dev#156).
+ → .cursor/skills/learnings/comment-line-width/
+
 ## toolchain
 
 

.cursor/skills/learnings/comment-line-width/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: comment-line-width
+description: >-
+  MUST be used when writing or reflowing code comments in this repo: wrap
+  at the rubocop line width of 120, not the legacy ~80 columns that most
+  existing comments still use.
+---
+
+# Comments wrap at 120, not the legacy 80
+
+Rubocop (via rubocop-shopify) sets `Layout/LineLength` Max to 120, but most
+comments predate the 80-to-120 expansion and are still wrapped at ~80 — and
+no cop flags a too-narrow comment, so the legacy style silently propagates
+by imitation. Fill new or edited comment blocks toward 120; the sweep of
+the legacy comments is tracked in d3mlabs/dev#156.
+
+Wrong:
+
+```ruby
+# Read as UTF-8 explicitly: the formulas have non-ASCII bytes (e.g. an
+# em-dash in a comment), and when release.rb runs under a non-UTF-8 locale
+# (such as a piped, login-less subshell) Ruby's default external encoding
+# is US-ASCII.
+```
+
+Right:
+
+```ruby
+# Read as UTF-8 explicitly: the formulas have non-ASCII bytes (e.g. an em-dash in a comment), and when release.rb
+# runs under a non-UTF-8 locale (such as a piped, login-less subshell) Ruby's default external encoding is US-ASCII.
+```
+
+learned-from: d3mlabs/dev#133 review — reviewer asked for the new
+bin/release.rb comments to fill the 120 width instead of copying the
+surrounding 80-column wrapping.
+date: 2026-09-10