Make the final self-check agree with the removal patterns

Step 3 treated any lost claim as an error, but patterns 2, 5, 6, 21,
25, 34, and 35 require removing claims, so the check rejected the
skill's own model answers and made behavior inconsistent. A lost
claim is now an error unless a pattern requires the removal, and
rule 2 asks to keep every supported claim. The em-dash
false-positive item now states it governs detection only, so it no
longer reads as an exemption from §14. Refs #212.

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

SKILL.md

@@ -21,7 +21,7 @@ The patterns below come from Wikipedia's ["Signs of AI writing"](https://en.wiki
 When given text to humanize:
 
 1. **Find AI patterns.** Check the text against the patterns below.
-2. **Keep every claim.** You may shorten dull parts, expand useful parts, and merge or split paragraphs. Keep the information even when you change the structure.
+2. **Keep every supported claim.** You may shorten dull parts, expand useful parts, and merge or split paragraphs. Keep the information even when you change the structure.
 3. **Do not invent facts.** Do not add a fact, name, number, date, quote, or citation unless it comes from the source or the user. If a sentence needs a missing detail, ask for it or use a simpler sentence. You may add an opinion or reaction when the writer's voice calls for one, but you may not add a factual claim. Fiction is exempt because invented details are part of the task.
 4. **Match the voice.** Use the right tone for the text, such as formal, casual, or technical. Add personality only when the text and the writer call for it.
 
@@ -403,7 +403,7 @@ A person may use some of these patterns. Do not treat any item below as proof by
 - **Letter-style opening or closing on a comment.** Salutations and sign-offs predate ChatGPT by centuries.
 - **Common transition words in isolation.** *Additionally*, *moreover*, *consequently* are AI-coded only when piled up. One *however* is not a tell.
 - **Curly quotes alone.** macOS, Word, Google Docs, and most CMSes auto-curl by default. Curly quotes only count when stacked with other tells.
-- **Em dashes alone.** Many editors and journalists use them often. Em dashes are evidence only when paired with formulaic sales-y rhythm.
+- **Em dashes alone.** Many editors and journalists use them often. Em dashes are evidence only when paired with formulaic sales-y rhythm. This item governs detection only: §14 still removes dashes from the final rewrite unless the writer's sample uses them.
 - **One short sentence for emphasis.** Flag dramatic fragments only when several appear in a row.
 - **Deliberate repeated openings.** Writers may repeat an opening to build rhythm or pressure, as in "She came. She saw. She conquered." Change it only when the repetition adds nothing.
 - **"Honestly" or "look" mid-sentence.** These are ordinary in casual writing. The tell is the standalone theatrical opener, not the word itself.
@@ -444,7 +444,7 @@ These details often carry the writer's voice. Keep them unless they hurt the mea
 3. Ask two questions:
    - **"What still sounds AI-generated?"**
    - **"Did the rewrite add or remove any fact, name, number, date, quote, citation, ranking, or other claim?"**
-   Treat any unsupported addition or lost claim as an error.
+   Treat any unsupported addition as an error. Treat a lost claim as an error unless a pattern above requires removing it, such as a name-drop list (§2), an unsourced claim (§5), a stock outlook (§6), a guess (§21), a send-off (§25), an unraised objection (§34), or a fake alternative (§35).
 4. Write the final version. State each point naturally instead of patching one flagged phrase at a time. If a sentence stays awkward, rewrite the paragraph around its main point. Apply the dash rule in §14.
 
 Return the result required by [How to return the result](#how-to-return-the-result).