Carolyn Bessette Kennedy: Biografi, død og Love Story

`.** The first characters of the input are a Danish preface – *”Her er den færdige HTML-artikel…”* – not `

`. (Gate: article-start)
2. **Placeholders `[…]` in the text.** Facts are incomplete in the source – e.g., *”John F. […]”* – and cannot be resolved without inventing content. (Gates: preserve facts, no fabrication)
3. **Embedded JSON instead of rendered HTML.** The input contains the `facts_json`, `ontology`, and `stats_line_plan` objects as raw text, including control strings like `{“claim”:”…”}` or `”label”: “Dødsdato”`. These are not valid rendered block content. (Gate: well-formedness)
4. **Meta-instruction text inside the document.** Lines like *”Required:

+

wrapper…”*, *”If language_code = “en”…”*, and *”Kun en enkelt Tier 2 (Yahoo, Newstimes)”* are directives, not article content. (Gate: valid content)
5. **Unresolved `”summary”` and `”faq”` parts are in JSON form,** not converted to HTML. (Gate: structure)
6. **Missing real values** such as the full date “7. januar 1966” is implied but not written; the name “John F. Kennedy Jr.” is truncated. These need to be factual, but they are missing from the source.

The provided document would first need to be reconstructed by the previous “P2” step (or the human editor) as a complete HTML string with all facts filled in and all non-HTML artifacts removed. Only *then* can the validator/repair layer correctly enforce the 13 gates.

If you can supply the clean, complete HTML article, I will run the mechanical validation and repair per the gates — moving the intro paragraph first, wrapping in `

`, inserting the stats line, normalizing snapshot cards, ensuring H2 sections end with a takeaway, placing callouts correctly, and outputting only the repaired HTML.