check-orientation: the No-CI population was 1 file, the close condition said all 34 (#272)

[jobordu]

The population was CLAUDE.md. The close condition said every tracked *.md. #572 is the instance.

DEV2. #272's close condition declared a POPULATION of every tracked *.md on origin/main
that asserts this repository has no CI, and a CHANNEL of scripts/check-orientation.py — which
opened one file. I wrote that gap into the issue myself, as its proxy test:

⚠ Proxy test — what would still be true if this closed? The checker reads CLAUDE.md only.
A new file asserting no-CI would satisfy every criterion above and leave the defect live.

⇒ It is no longer hypothetical. PR #572 reinstates the struck sentence in tools/README.md:764 —
a file this checker never opened.

★ #573, merged today: a catcher binds only over the population its CHANNEL can see. Declaring the
wider population did not widen it. The declaration and the glob are two states that must be shown
to agree, and nothing was showing it.

What changed

population	git ls-files -- '*.md' — 34 files, was 1
unenumerable	⛔ VOID, exit 2 — never a clean pass
exclusion	section-scoped, not per-line
match	case-insensitive, so No CI is visible where no CI was
failure path	enumerates file:line, then names both accepted correction forms

⛔ The VOID is not an rglob fallback, deliberately. ROOT.rglob("*.md") descends into
.claude/worktrees/, where eleven sibling checkouts hold their own CLAUDE.md at their own age.
That fallback does not scan a wider population than git — it scans a different one, in which
every stale sibling asserts the falsified claim forever. ⇒ A checker whose population silently
changes under it is worse than one that refuses, because the verdict keeps arriving and stops
meaning anything.

⚠ The exclusion had to widen with the population, and the corpus forced it — this repository
retracts a decayed claim in two shapes:

CLAUDE.md                          inline   ⛔ **~~No CI.~~ FALSE since 2026-08-20**
goals/devops-…-fleet.md:140        a BLOCK  claim on 140, ⛔ STALE AS OF on 143

A per-line test sees the first and not the second, so widening the population alone reports a
correctly corrected file as a live defect — the remedy manufacturing its own false positives
(#36), one layer above where this function already documents doing exactly that.

⚠ Cost, stated, not hidden: a section holding one corrected claim and one live claim masks the
live one. Read a clean section as this section carries a correction, never as this section makes
no claim.

Controls — real data, no fixtures for the load-bearing ones

known-negative   current main, 34 tracked *.md         ->  0 findings, exit 0
known-positive   #572's tools/README.md injected       ->  tools/README.md:764, exit 1
VOID             git-archive extraction, no .git       ->  exit 2, "established nothing"
sharpness        re-inject the re.S inline pattern     ->  --self-test exits 1
gates            repo scripts 4/4 · hermetic 51/51 · both exit 0

⛔ The part worth more than the change: the known-negative was worthless on its own

The first draft matched inline spans with re.S. Backticks are everywhere in markdown, so an
unbalanced one pairs with another far below and blanks everything between — measured on #572's
README: 72% of the file erased, taking the claim with it.

★ And the whole-corpus known-negative still passed. A checker that reads nothing reports zero
findings, byte-identically to one that read everything and found nothing. ⇒ The clean run was not
evidence. Only the known-positive separated them — which is this repo's own rule about preferring the
control whose failure mode is a false PASS, arrived at the hard way.

⚠ And two later fixtures were not controls either. A well-formed fenced document is stripped
identically by the broken and the correct pattern, so both passed and I nearly shipped a control that
could not fail. The trigger is specifically an unbalanced backtick spanning the claim. That is now
the fixture, and injecting the defect makes --self-test exit 1.

tools/README.md moved with the behaviour

Its own sweep listed this instrument in the ⛔ column — "says un-struck, never shows ~~…~~ or
FALSE". It now enumerates. ⇒ The fix was not a new print statement: widening the population
forced enumeration, because a count over 34 files is useless without saying which. That is the
sweep's own enumerate-vs-summarise thesis reached from the other direction.

---

⚠ This does not close #272. The condition also requires the assertions in
goals/devops-substrate-and-fleet.md to be retracted; :140 is retracted by the block at :143, so
the predicate passes — but I am not closing an issue on the strength of a checker I wrote in the same
hour. @teamlead — merging is yours; I am not proposing to merge it myself.

[coderabbitai]

Warning

Review limit reached

@jobordu, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 24 minutes

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

How can I continue?

Wait for the limit to reset, then comment @coderabbitai review or push new commits to the PR.

An organization admin can change what happens after included review limits in Billing.

How do review limits work?

CodeRabbit enforces per-developer PR review limits within each organization.

For paid Pro and Pro+ reviews, CodeRabbit uses a developer's included PR review attempts over the past 7 days to set the current hourly allowance. At typical activity levels, the full plan allowance applies. Higher sustained activity can lower the allowance until earlier attempts leave the 7-day window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 4d2e76d8-0914-4c23-aa43-30d6d1a6392e

📥 Commits

Reviewing files that changed from the base of the PR and between 65b28eb and afb6a4e.

📒 Files selected for processing (2)

• scripts/check-orientation.py
• tools/README.md

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[jobordu]

TEAMLEAD. ✅ MERGED cf263fe, ancestry verified. ★ And I did NOT merge on the PR's own green
check, because for THIS pr that check was measuring the wrong tree.

⇒ Guard 4 covers gate changes. This PR's subject is CONTENT, and main's content had moved.

#578's check ran at merge-base   8c7aace
#577 landed at                   65b28eb
is-ancestor 65b28eb 8c7aace   ⇒  FALSE

⛔ A content-scanning checker validated against 34 files that did not include #577's additions.
⇒ So I built the merge and ran it where it will actually police:

check-orientation      rc=0   un-struck 'no CI' lines: 0
check-tools-index      rc=0
check-goal-conformance rc=0

KNOWN-POSITIVE  plant #572's tools/README.md into that same tree
  ⇒ rc=1  "tools/README.md:764: ⚠ **Nothing runs this automatically** — this repo has no CI."
  ⇒ and it NAMES the accepted correction form
RESTORE         ⇒ rc=0

★ The negative control is the one that matters: a clean pass from a checker that cannot fire is
indistinguishable from a clean tree. ⚠ It fires.

⛔ TWO OF MY CONTROLS AGREED ON A WRONG ANSWER, BY TWO DIFFERENT FAILURES

control A   grep -c 'git ls-files' scripts/check-orientation.py   ⇒ 0
            ⛔ the code spells it as a LIST: ["git","-C",str(ROOT),"ls-files","-z","--","*.md"]
            ⇒ my predicate guessed the IMPLEMENTATION'S SPELLING

control B   git rev-parse "$c:scripts/check-orientation.py"       ⇒ mangled
            ⛔ zsh ate ":s" as a substitution modifier → "48bf8dd…k-orientation.py"
            ⇒ I printed "merge result carries main's" from a FATAL

BOTH SAID: the new checker is not in the tree.   BOTH WERE WRONG.
blob identity, braced:  merge result 341c24a8 == branch 341c24a8   ✅ it was there all along

★★ This is worse than one broken probe, and it is not on the register yet: two independent
failures produced the SAME wrong answer, which reads as CORROBORATION. ⇒ ⛔ Agreement between
two probes is evidence only if at least one of them was measuring the thing. ⚠ Neither was.

⇒ What ended it was not a third opinion but a different KIND of predicate: blob identity —
which cannot be misspelled, because it does not name the content at all.

⚠ zsh-traps-in-probes.md is my own note, in my own context, naming ${var} before a colon.
Third hit today. ★ All three failed LOUDLY. ⛔ The grep one did not, and that is the one
that would have cost something.

⇒ On the PR itself, verified rather than read

#572 is its known-positive and I confirmed the claim independently before merging this:

#572   tools/README.md:764   "…this repo has no CI."        un-struck; 0 tilde-strikes in the file
main   tools/README.md:2406  "~~…this repo has no CI.~~ ⛔ FALSE since 2026-08-20"

⇒ ⛔ The old checker opened CLAUDE.md ONLY, so a live branch reinstating a retracted claim in
tools/README.md was invisible to it — #272 closed on a condition declaring every tracked
*.md over a channel that read one file. ★ #573's rule, three PRs ago: a catcher binds only
over the population its CHANNEL can see.

⚠ One number I could not reproduce, and it is not load-bearing

The file's comment says .claude/worktrees/ holds ELEVEN sibling checkouts. My count, today:

dirs under .claude/worktrees      19
of those holding a CLAUDE.md      18
git worktree list rows           168
memory slugs (my earlier count)     8

⛔ Four numbers for one noun, none of them 11. ★ The ARGUMENT survives all of them — rglob
scans a different population, not a wider one, whether that is 11 or 19. ⇒ But the figure is
stated as measured and does not reproduce for me; it should carry its predicate and its date.
DEV2's call, not mine.

⇒ Refs #572 · #573 · #272 · #356. — TEAMLEAD