docs: R-302 — correcting R-300, the pathspec was not the fault

R-300 blamed `git add -A` for eead663 carrying my uncommitted edits.
check-figures' author checked and it was not that: they staged two exact
paths. The tree agrees — eead663 holds CLEAN_GROUND.md and CLEAN_GROUND_ART.md
only, while my modified step5-split.mjs and untracked check-unmarked.mjs are
absent, both of which `git add -A` would have taken.

`git add <file>` takes the whole file including another session's edits to it,
so a pathspec stops you sweeping files you did not touch and does nothing about
the one you did. Their pre-commit `npm run check` passed because my uncommitted
module was in the shared tree: the commit was broken, the working copy was not.

I asserted a mechanism I never looked at, in an entry whose technical finding
was correct. The correction came from the session it accused.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/REVIEW_LOG.md

@@ -7363,3 +7363,39 @@ obvious, and two readers that disagree are how three of this session's defects w
 
 **State:** CLEAN GROUND v0.24, **22 guards**, 168 cited figures, artwork prompts written and
 nothing generated. Still the only untested thing is a run with human beings.
+
+---
+
+## R-302 — correcting R-300: the pathspec was not the fault
+
+**R-300 says `eead663` swept my uncommitted edits in by `git add -A`. That is wrong, and the
+truth is worse.** check-figures' author staged two exact paths, which is the rule this repo
+has been following all session. I did not check before writing it down; they did, and then
+told me. The tree agrees with them — `eead663` contains `CLEAN_GROUND.md` and
+`CLEAN_GROUND_ART.md` and nothing else, while my modified `tools/step5-split.mjs` and
+untracked `tools/check-unmarked.mjs` are absent from it. `git add -A` would have taken both.
+
+**`git add <file>` takes the whole file, including the edits another session made to it.** Two
+sessions had touched `CLEAN_GROUND.md`; eight of that commit's eleven hunks were mine. A
+pathspec names a file, not a change, so **staging exact paths stops you sweeping files you did
+not touch and does nothing about the file you did.**
+
+**And the green build was green for the wrong reason.** `npm run check` passed for them before
+committing, because my uncommitted `step5-split.mjs` was sitting in the shared tree. The
+commit was broken; the working copy was not. That is this log's oldest theme arriving in the
+build system itself — **a check that passes for a reason unrelated to what it claims** — and
+it is the reason `check-cited` failed on a fresh clone of HEAD while passing for everyone who
+had the file on disk.
+
+**Two rules, both needed, neither sufficient alone:**
+
+- **`git diff --cached` before committing.** The hazard is the file, not the pathspec, and the
+  staged diff is the only thing that shows what is actually going in.
+- **Prove the commit standalone.** Clone HEAD into a tempdir and run the suite there, or
+  `git checkout-index -a --prefix=`. A green run in a shared working tree says nothing about
+  what you pushed, because the tree contains the thing the commit is missing.
+
+**Worth recording that the correction came from the person it accused.** They reproduced my
+finding before accepting it, repaired it in `e9f7d45`, then checked my account of their own
+commit and found it wrong. R-300's technical claim stood; its account of how the breakage
+happened did not, and I had asserted a mechanism I never looked at.