P3 [CRF-6] The PR description's verification section says the rule "flags the 7 instances the cleanup commit fixes." Vale scopes to [*.md] only (.vale.ini line 47); it can only flag the 5 Markdown instances. The 2 manifest.json fixes are manual edits outside Vale's scope. The claim conflates Vale-flagged instances with manually-fixed entries. The correct statement is: Vale flags 5 instances; 2 additional instances in manifest.json are fixed by hand. (Mafu-san)

🤖

P3 [CRF-4] "cleanup landed in the previous commit" is a temporal reference that dissolves after merge. After a squash merge, "the previous commit" doesn't exist. After a rebase merge, the SHA changes. The factual claim (violation count is zero) stands on its own. Drop the parenthetical:

# count is zero, so this rule can ship at error from day one.

(Leorio)

🤖

P2 [CRF-2] The comment says "the %s token interpolates the matched (wrong) text" (singular). The message template has two %s tokens with different semantics: the first interpolates the correct (replacement) form, the second interpolates the matched (wrong) form. A contributor adapting this template or debugging a message that reads backward gets a wrong mental model.

Suggestion:

# unchanged; the first %s interpolates the correct (replacement) text
# and the second interpolates the matched (wrong) text.

(Leorio)

🤖

P3 [CRF-7] The header comment is 17 lines for 11 lines of config. Roughly half restates identifiers or duplicates content from style-guide.md:

• Line 1 (Coder.BrandNames - enforce canonical brand-name casing) restates the filename and extends: substitution.
• Lines 3-7 (substitution scope) duplicate style-guide.md:47.
• Lines 15-17 (how to add a brand) duplicate style-guide.md:56-58.

The trap (false-positive scope) and why-not-what (error level) carry their weight. Trimmed draft that keeps both:

# substitution rules skip code blocks, inline code, and URLs, so
# provider sources (hashicorp/kubernetes) and URLs are unaffected.
#
# Error from day one: brand casing is objectively right or wrong,
# and existing-content violations are zero.
#
# Adding a brand: append to swap, audit docs for violations.

(Gon)

🤖

Note [CRF-12] Worth knowing: make lint/prose uses --no-exit (Makefile:854), so Coder.BrandNames at error level surfaces as a CI annotation but does not fail the build or block merge. The enforcement depends on human reviewers noticing the annotation. This is a system-level constraint from the 391 pre-existing errors in upstream Google rules, not a flaw of this PR. (Pariston)

🤖

P4 [CRF-8] No regression fixture for the rule. If someone later breaks the swap YAML or removes an entry, there's no automated check that the rule still fires on known-bad input. The "test" is the absence of violations in the corpus, which is indistinguishable from a broken rule that fires on nothing. Not this PR's debt (no Vale rule test infrastructure exists in the repo), but worth noting as the Coder/ style directory grows. (Bisky)

🤖

P3 [CRF-5] The swap table catches Hashicorp and HASHICORP but omits hashicorp (all lowercase). The PR adds HASHICORP defensively despite zero corpus occurrences, but omits hashicorp citing 49 URL/code instances. However, this rule extends: substitution, and Vale's substitution scope already skips fenced code blocks, inline code, and URLs by default (as the comment on lines 3-7 documents). Adding hashicorp: HashiCorp would catch the one case the current table misses (lowercase in prose, which is always wrong for a proper noun) with zero false-positive risk.

The stated fallback (Vale.Spelling, DOCS-187) doesn't exist yet. Until it ships, a contributor who writes "hashicorp" in body text gets no lint signal.

swap:
  hashicorp: HashiCorp
  Hashicorp: HashiCorp
  HASHICORP: HashiCorp

(Zoro P3, Hisoka Note)

🤖