Step 4's COMPILE_CMD and LINT_CMD rows state a site count in prose that nothing checks

[dmccoystephenson]

Summary

The COMPILE_CMD and LINT_CMD rows of the Step 4 substitution table describe their substitution sites by count rather than by name:

• COMPILE_CMD (create-dev-loop.md:624): "Substitute the same text in both places it appears (Phase 3 verify block, Phase 8 rebase fence)."
• LINT_CMD (create-dev-loop.md:626): "Appears in both the Phase 3 verify block and the Phase 8 rebase fence — include or omit it in both".

Both are accurate today (two occurrences each in the template body, verified by counting {{COMPILE_CMD}} and {{LINT_CMD}} tokens between the outer fences). The TEST_CMD row used the same "both places" phrasing until PR #151, and it went stale the moment PR #121 added a third occurrence — the word "both" encodes a count that nothing checks, so a later PR that adds a site has no prompt to revisit the row.

Why it matters

The drift mechanism recorded in #139 is not specific to TEST_CMD: any row that states a site count in prose will silently disagree with the template after an occurrence is added. scripts/check_docs.py verifies that a row exists per placeholder, not that its prose matches the template's occurrence count, so the disagreement passes CI.

Suggested directions

Offered as options for the maintainer:

1. Drop the count words. Rephrase both rows to "Substitute the same text everywhere {{COMPILE_CMD}} appears (Phase 3 verify block, Phase 8 rebase fence)", matching the TEST_CMD row's form after Enumerate all three TEST_CMD substitution sites in the Step 4 row #151. The parenthetical can still go stale, but the sentence no longer asserts a number that contradicts the template.
2. Make the count mechanical. Add a check_docs.py case that, for each placeholder, compares the number of occurrences in the template body against a machine-readable annotation on the row (e.g. a trailing <!-- sites: 2 --> comment), with a fixture case in tests/test_check_docs.py that fails when they disagree. This is the only option that catches the class rather than the instances, at the cost of a new convention in the table.

Option 1 is a two-line docs fix; option 2 is a small Stage B / check-expansion cycle. Either is appropriately scoped for one loop cycle.

Research grounding

No RESEARCH.md finding applies. This is a documentation-accuracy observation with no empirical claim behind it.

Provenance

Observed as an out-of-diff finding during the Phase 4 self-review of PR #151, which corrected the TEST_CMD row. Rewording the sibling rows was outside #139's scope.

This issue body was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson