fix(gip): spell the base branch as <default-branch> throughout

[d-morrison]

Closes the one item PR #1008 deliberately left behind: skills/gip/SKILL.md
gave the same instruction two ways.

#1008 parameterized the base branch to origin/<default-branch> throughout the
worktree-isolation steps, and correctly Deferred the remaining literal
main spellings rather than fixing them on that branch. They were
byte-identical to main at the time, so editing them there would have put the
branch out of step with main over content it did not own -- the main-sync
rule in address-every-comment.md, and the authorship-not-adjacency line in
dont-incur-technical-debt.md.

That rule's stated follow-up is "fixing it on main directly, not a new
issue". #1008 merged as 3eb15a4, so that is now possible, and this is it.

What changed

Line	Before	After	Why
40	not main	not <default-branch>	prescriptive: names the base a later issue's branch comes from
71	branch straight from origin/main	branch straight from origin/<default-branch>	prescriptive: the independence gate's own criterion

Line 71 also carried an em-dash. Editing a line makes its pre-existing
violations part of this diff, per ascii-punctuation-in-source.md, so that one
is now --. The file's other em-dashes are untouched, since a whole-file
replace is a different and much larger change than the finding asks for.

What was deliberately left alone

Two occurrences of main remain, and both are evidence rather than
instruction:

• L120 -- the sentence naming "the literal string main" as the thing not
  to hard-code. It is about the string, so substituting it erases the point.
• L121 -- the verbatim fatal: invalid reference: origin/main. A reader
  matches that against their own terminal output, so a parameterized version
  matches nothing.

The sweep was mechanical rather than by eye, because the recurrence had been
under-reported at earlier rounds: a word-boundary scan for main over the
whole file returned exactly four occurrences, and each was classified before
anything was edited. Four before, two after.

The general rule behind that split -- a consistency edit must stop at anything
asserting what was observed -- is recorded separately in #1014, which found it
unwritten anywhere in the corpus even though #1008 had applied it correctly.

Verification

• Word-boundary \bmain\b sweep over the file: 4 occurrences before, 2 after,
  both remaining ones intended.
• scripts/sync-codex-skill-wrappers.py re-run: 175 wrappers regenerated, no
  drift (git status clean afterward).
• Banned-glyph scan over git diff -U0 origin/main...HEAD (three-dot range,
  run after committing, LC_ALL=C.UTF-8): 5 added lines examined, 0 hits, 0
  non-ASCII of any kind.
• NLB_BASE_REF=origin/main check-new-line-breaks.py: "No lines missing
  semantic breaks."
• Push confirmed landed by git ls-remote origin fix/gip-default-branch-spelling
  matching git rev-parse HEAD at f368be4.

Disjoint from #1014, which touches only shared/ fragments. No merge-order
constraint between them; either can land first.

---

Generated by Claude Code

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.

_{2026-08-01 05:37 UTC}

[claude]

Code review

Reviewed skills/gip/SKILL.md (the only file changed): two remaining main spellings in the independence-gate/base-branch-stacking prose are parameterized to <default-branch>, and the pre-existing em-dash on the line touched in the second hunk is converted to --, consistent with ascii-punctuation-in-source.md's editing-a-line-makes-its-pre-existing-violations-yours rule.

Verification performed:

• Confirmed via origin/main's prior content that the file had exactly 4 word-boundary main occurrences before this PR (lines 40, 71, 118, 119) and exactly 2 after (lines 120, 121 in the new file), matching the PR body's claim.
• Confirmed the two remaining occurrences are evidence, not instruction: one names the literal string main itself, the other is the verbatim fatal: invalid reference: origin/main error text a reader matches against their own terminal; both are correctly left alone.
• Confirmed both cited fragments exist verbatim: shared/workflow/address-every-comment.md's main-sync-case language (line 394) and shared/principles/dont-incur-technical-debt.md's authorship-not-adjacency line (line 113); no hallucinated citations.
• Confirmed PR ums: three findings from a five-PR session #1008 merged as 3eb15a4 per the PR body's premise (verified against gh pr view 1008).
• Confirmed the PR head SHA (f368be4...) matches the live headRefOid.
• No stale duplicate of the old origin/main-in-prose phrasing found elsewhere in the corpus, including the generated codex-skills/gip/SKILL.md wrapper, which only points at the canonical source and carries no duplicated content.
• Rest of the file already uses <default-branch> consistently, so the change matches the file's own established convention; markdown list formatting and line-wrapping remain valid after the edit.

No bugs, no CLAUDE.md/lab-manual violations, and no hallucinations found. This is a minimal, well-scoped, and thoroughly self-verified documentation fix.

Verdict

Ready for merge

[github-actions]

Claude finished review — View run

---

Posted the review comment successfully: #1013 (comment)

Summary

Reviewed PR #1013, a minimal fix to skills/gip/SKILL.md parameterizing two remaining main spellings to <default-branch> and fixing a pre-existing em-dash on one of the touched lines.

Verification performed:

• Confirmed the file had exactly 4 \bmain\b occurrences before (vs. origin/main) and 2 after — matching the PR's stated sweep.
• Confirmed the 2 remaining main occurrences are legitimate (naming the literal string, and a verbatim error message) rather than missed instances.
• Confirmed both cited fragments (address-every-comment.md's main-sync case, dont-incur-technical-debt.md's authorship-not-adjacency line) exist verbatim — no hallucinated citations.
• Confirmed PR ums: three findings from a five-PR session #1008's merge SHA (3eb15a4) and this PR's head SHA (f368be4...) as stated.
• No stale duplicates elsewhere in the corpus (including the generated codex-skills wrapper).

No bugs, no CLAUDE.md/lab-manual violations, no hallucinations found.

Verdict

Ready for merge

[github-actions]

💰 Cost: $4.5594 (review) — run

[d-morrison]

Ready for merge, on claude-review's verdict alone.

All 10 check runs completed green at f368be4, including require-review, and the review bot returned Ready for merge with no findings and zero inline review threads.

copilot-pull-request-reviewer completed success at 05:30:47Z having posted nothing --- zero formal reviews, zero review comments on this PR. That is the third state of that reviewer recorded in #1014 (green check, no review, and no refusal either), so it is not evidence of agreement. The clean call rests on one reviewer, and this is a live instance of the case #1014 is documenting while #1014 itself is open.

Verified here rather than relayed from the review: git show origin/main:skills/gip/SKILL.md | grep -nE '\bmain\b' returns 4 occurrences (lines 40, 71, 118, 119), and the same command against the head returns 2 (lines 120, 121) --- both of them the evidence cases the body names. The 4-before/2-after sweep holds.

---

Generated by Claude Code