A list of data is not a tricolon

SKILL.md

@@ -128,6 +128,7 @@ respecto a, de cara al futuro, en este artículo
   signpost. · "en conclusión", "en resumen".
 - **Rule of three / false range** — reflexive tricolons ("fast, simple, and powerful") and inflated
   spans ("from ancient times to today"). Vary the count; keep a range only if the middle matters.
+  A list of data is not a tricolon: leave "2324, 2425 and 2526" or "CD31, CD45 and CD56" as written.
 - **False balance** — "on one hand… on the other" when the evidence favors one side. Say which.
   · "por un lado… por otro".
 

src/SignsOfAI.Core/Rules/Packs/rules.en.json

@@ -888,7 +888,7 @@
       "id": "rhet.rule-of-three",
       "humanRatePer1000": 1.09,
       "category": "Rhetorical",
-      "regex": "\\b(\\w+), (\\w+),? and (\\w+)[.,;]",
+      "regex": "\\b(?!(?-i:[\\p{Lu}\\d]+), (?-i:[\\p{Lu}\\d]+),? and (?-i:[\\p{Lu}\\d]+)[.,;])(\\w+), (\\w+),? and (\\w+)[.,;]",
       "weight": 1.2,
       "severity": "Info",
       "message": "Possible “rule of three”: AI defaults to tricolons.",

src/SignsOfAI.Core/Rules/Packs/rules.es.json

@@ -758,7 +758,7 @@
       "id": "rhet.regla-de-tres",
       "humanRatePer1000": 1.29,
       "category": "Rhetorical",
-      "regex": "\\b(\\w+), (\\w+) y (\\w+)[.,;]",
+      "regex": "\\b(?!(?-i:[\\p{Lu}\\d]+), (?-i:[\\p{Lu}\\d]+) y (?-i:[\\p{Lu}\\d]+)[.,;])(\\w+), (\\w+) y (\\w+)[.,;]",
       "weight": 1.2,
       "severity": "Info",
       "message": "Posible “regla de tres”: la IA abusa de los tríos.",

tests/SignsOfAI.Core.Tests/RuleOfThreeTests.cs

@@ -0,0 +1,69 @@
+using SignsOfAI.Core;
+using Xunit;
+
+namespace SignsOfAI.Core.Tests;
+
+/// <summary>
+/// The rule-of-three tell is a <em>rhetorical</em> tricolon: three words chosen for cadence, "fast,
+/// simple, and powerful". Its regex matched any three word-characters, so it could not tell that
+/// cadence from an author listing three values — financial-aid year codes, gene names, the numbers
+/// of a dose schedule. #100 found it in a support-ticket reply, then measured it: 18 of the rule's
+/// 236 matches in the calibration corpus are lists like that, and six texts had no other hit.
+///
+/// The narrowing lives in the pack, not in C#, so the whole rule stays editable by anyone: a list is
+/// skipped when none of its three items contains a lowercase letter. Two traps shaped the regex, and
+/// both are pinned below — the analyzer compiles every pack pattern with IgnoreCase, which would make
+/// an uppercase class match prose too, and <c>[A-Z]</c> would miss an accented acronym.
+///
+/// It changes no published number: the calibration run before and after is identical in every field.
+/// </summary>
+public class RuleOfThreeTests
+{
+    private readonly AiWritingAnalyzer _a = new();
+
+    private bool Fires(string text, string lang, string ruleId) =>
+        _a.Analyze(text, lang).Findings.Any(f => f.RuleId == ruleId);
+
+    // The seven published-paper lists #100 listed from the corpus, and the two from the ticket.
+    [Theory]
+    [InlineData("Expression was measured for DPM1, DPM2 and DPM3.")]
+    [InlineData("Cells were stained for CD31, CD45, and CD56, as described.")]
+    [InlineData("The panel covered ARS, RQR and RRS.")]
+    [InlineData("Exposure to IMI, CTD, and THX, alone or combined.")]
+    [InlineData("Group sizes were 6, 6 and 7, respectively.")]
+    [InlineData("Samples were taken at weeks 25, 26, and 28.")]
+    [InlineData("Doses of 10, 15, and 20.")]
+    [InlineData("The same is true of 2324, 2425 and 2526.")]
+    [InlineData("It drops SENIOR, COMMUNITY, COMM, CMTY and CHD, and matches the rest.")]
+    public void A_list_of_data_is_not_a_tricolon_en(string text) =>
+        Assert.False(Fires(text, "en", "rhet.rule-of-three"), text);
+
+    [Theory]
+    [InlineData("The tool is fast, simple, and powerful.")]
+    [InlineData("We invited Alice, Bob and Carol.")]
+    // Mixed: one prose item makes it a sentence again. Requiring all three to be data is the
+    // conservative choice #100 argued for, and the one its numbers measured.
+    [InlineData("They expressed CD31, CD45, and cells.")]
+    public void A_rhetorical_tricolon_still_fires_en(string text) =>
+        // If IgnoreCase leaked into the token class, lowercase prose would read as data and these
+        // would stop firing. That is the trap the inline (?-i:) exists to avoid.
+        Assert.True(Fires(text, "en", "rhet.rule-of-three"), text);
+
+    [Theory]
+    [InlineData("Se aplicó en los cursos 2324, 2425 y 2526.")]
+    [InlineData("Participaron la UE, la OTAN y la ONU; también UE, OTAN y ONU.")]
+    // Accented acronyms: \p{Lu} rather than [A-Z], because the Turkish dotless i taught this project
+    // what an ASCII-only letter class does to other alphabets.
+    [InlineData("Se revisaron ÁREA, ÉPOCA y ÍNDICE.")]
+    public void A_list_of_data_is_not_a_tricolon_es(string text) =>
+        Assert.False(Fires(text, "es", "rhet.regla-de-tres"), text);
+
+    [Theory]
+    [InlineData("La herramienta es rápida, sencilla y potente.")]
+    // Single-word items, as the rule has always required: with articles ("el área, la época") it never
+    // matched, before or after this change. Lowercase accented letters must still read as prose.
+    [InlineData("Revisaron área, época y índice.")]
+    [InlineData("Acordaron UE, OTAN y aliados.")]
+    public void A_rhetorical_tricolon_still_fires_es(string text) =>
+        Assert.True(Fires(text, "es", "rhet.regla-de-tres"), text);
+}