ums: re-audit after programmatic comment edits; sibling-merge clobber hazard

[d-morrison]

Banks two recurring lessons from the sparta #143/#150/#152 arc into memories/debugging.md.

What

1. Programmatic comment edits leave punctuation/grammar artifacts. Scripted removal of issue refs (and parentheticals/clauses) from comments repeatedly broke the surrounding prose — orphaned commas/periods, comment-marker-plus-bare-punctuation (##.), dangling "see issue.", an exposed repeated word. The reviewer/Copilot flagged ~6 of these across #150 and #152. The note lists the failure shapes, the fix for each, and the audit greps — plus which sed patterns strip cleanly vs which cases need hand edits.

2. Merging main into a sibling PR can silently clobber an un-customized file. When one PR merges and you sync a sibling that touches the same files, a file the sibling never customized takes main's version with no conflict — so it can end up describing the wrong change. Hit on ums: add rme/Epi 204 working-notes memory #152, where demos/demo.json silently became Genericize incidental project-specific references in skills #143's diplomacy reason. The note says to diff the whole merge result vs the PR's intent (not just resolve marked conflicts) and re-run the PR's own invariant over files main re-touched.

Why

Both bit this session more than once and weren't recorded anywhere; encoding them should pre-empt the same review rounds next time.

🤖 Generated with Claude Code

---

Generated by Claude Code

[Copilot]

Pull request overview

This PR updates memories/debugging.md to capture two recurring review/debugging lessons from the sparta scrub PRs: (1) scripted comment edits can leave behind punctuation/grammar artifacts, and (2) merging main into a sibling PR can silently overwrite (“clobber”) files that weren’t actually customized in the sibling.

Changes:

• Adds a checklist of common punctuation/grammar failure modes after programmatic comment scrubs, with suggested audits/greps.
• Adds guidance on verifying merge results when syncing a sibling PR after another PR lands, to catch silent “clean merges” that change PR-specific files.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[claude]

Claude finished @d-morrison's task in 2m 51s —— View job

---

Review complete

• Gather context
• Read diff vs origin/main
• Review current state of memories/debugging.md
• Check for correctness, AI tells, hallucinations
• Post findings

---

No new findings. All prior inline findings have been addressed, and the latest commit (62fc436) correctly resolves the rev-parse HEAD imprecision flagged in the previous round.

Technical verification:

• grep -rnE "^[[:space:]]*#+[[:space:]]*[.,;:]" — Valid ERE; POSIX char classes work with -E; the pattern correctly fires on lines like # , or ##.. ✓
• s/ \(#[0-9]+\)//g — In sed -E, \( and \) match literal parens (ERE groups use unescaped ()), so this cleanly strips (#NNN). ✓
• s/^# #[0-9]+: /# / — ^-anchored, no g. Correct. ✓
• s/, #[0-9]+,/,/g — Strips ref between commas, leaving one comma. Correct. ✓
• git log --oneline HEAD..origin/<branch> — Lists commits reachable from origin/<branch> but not from HEAD. Empty when origin is where you left it (regardless of unpushed local commits); non-empty when a parallel push moved origin past you. The parenthetical correctly explains why the bare rev-parse equality check fails in the "have local commits" state. ✓

Cross-repo refs: All three sparta references use fully-qualified form (Lacaedemon/sparta#143, Lacaedemon/sparta#150, Lacaedemon/sparta#152). ✓

AI tells: None. Prose is specific, concrete, and direct throughout. ✓

Hallucinations: None. GitHub's "Update with rebase" button is real; all git commands describe accurate behavior. ✓

This PR is clean and ready to merge.