n-tv (deutscher […]
| Gesundheitslage 2026 |
„sehr gut“ nach eigenen Angaben (n-tv (deutscher Selbstbild„Ich bin ein zähes Luder“ (n-tv (deutscher AuszeichnungVerdienstorden des Landes Nordrhein-Westfalen 2026 (` maybe with “? Not specified. Need close row/table. We can reconstruct:
Serienende2020 (Stern (deutsches Nachrichtenmagazin)td> |
Need maybe include “Zweite Diagnose” row. The snippet “Morbus Hodgkin (n-tv (deutscher […]
| Gesundheitslage…” Missing ` |
` before next row. We’ll repair.
5. `
Georg Uecker lebt nach Angaben von n-tv (deutscher Nachrichtensender) seit 1993 mit HIV. Im Zusammenhang mit dieser Diagnose wurde bei ihm auch Morbus Hodgkin festgestellt – eine Krebserkrankung des Lymphsystems, wie n-tv (deutscher Nachrichtensender) berichtet.
`
– This is complete.
6. `
Schon 2018 beschrieb Uecker im Interview mit der taz (überregionale Tageszeitung), dass keine messbaren Viren mehr […] Damals sagte er, er sei vermutlich schon weit über 30 Jahre HIV-positiv, wie TV […] im Zusammenhang damit wird Morbus Hodgkin festgestellt (n-tv (deutscher […]`
– This is broken. Need reconstruct sentence:
“Schon 2018 beschrieb Uecker im Interview mit der taz, dass keine messbaren Viren mehr [nachweisbar seien]. Damals sagte er, er sei vermutlich schon weit über 30 Jahre HIV-positiv, wie TV Spielfilm berichtete. Im Zusammenhang damit wird Morbus Hodgkin festgestellt (n-tv).”
But we must not invent. The raw has “dass keine messbaren Viren mehr […] Damals sagte er, …”. The ellipsis might indicate omitted text “nachweisbar seien” but not in raw. We cannot invent. We can output incomplete? No, raw article likely had more. The prompt may have truncated with “[…]” for brevity. We should not invent missing words. But to repair HTML, we need valid sentence. We can maybe leave “mehr […]” as is? That would be factually incomplete. But user says “Do NOT rewrite article or change facts. Only repair per gates.” If original P2 output had ellipses? Maybe not; the prompt’s ellipses are from truncation by the user? Hmm.
Need decide. The “raw article (P2 output)” in the code block likely contains actual raw output, not abbreviated. But the text includes “[…]” because maybe the platform’s context got truncated? Let’s look at exact: “dass keine messbaren Viren mehr […] Damals sagte er” There is no closing period. It might be a “placeholder” inserted by the system after truncating due to token limit. Similarly many `[…]` throughout. So we are not given full raw article. We need maybe repair only the visible fragments? Impossible to produce a comprehensive article without inventing.
Maybe the task is not to reconstruct entire article, but to “validate + repair layer” by outputting a list of repairs? Actually if raw article is truncated, perhaps we should say cannot fully repair due to missing content? But the user expects output maybe from a machine. Need be helpful.
Let’s inspect the JSON at the end. It includes claims and source info. Could be a “facts” array from a fact-checking pipeline. The user might want us to validate the article and output JSON with fact-check results. But the role says “Validator + repair layer for WordPress-ready HTML.” It doesn’t say “fact-check.” The JSON contains claims and “must_cover_claims” maybe from another part. Maybe we need not worry.
Let’s identify the “13 gates” from context? Could be:
1. Article must have “ tag.
2. First child of article must be intro p with class and itemprop.
3. Intro p max 2 sentences.
4. No title/h1 in article? (WordPress already has title)
5. Links target=_blank rel=noopener nofollow.
6. Source anchor text should include source name and descriptor in parentheses.
7. No broken HTML.
8. Use “ for FAQ.
9. If language_code = “en”, leave labels as-is.
10. Don’t invent facts.
11. Repair missing tags.
12. Preserve facts.
13. Maybe “must_cover_claims” from JSON must be included.
But not sure.
Let’s see if the JSON includes “must_cover_claims”: “Er erhielt 2026 den Verdienstorden NRW”, “Er engagiert sich für die queere Community”, “Tritt gelegentlich als Autor und Schauspieler auf”. These claims are not in raw visible article except first. “Engagiert sich für queere Community” and “Tritt gelegentlich als Autor und Schauspieler auf” might need to be covered? But raw article has no mention of queere Community or Autor. User said “must_cover_claims” maybe part of fact-check requirements. If we are repairing article, do we need add these facts? No, “Do NOT rewrite article or change facts.” Maybe the article P2 output should have covered them but didn’t? The instruction says “must_cover_claims” maybe from a “ClaimCoverage” gate requiring the article to cover claims. But “Only repair per 13 gates below” might include “must_cover_claims” as a gate? Hmm.
Let’s think of a possible hidden system prompt. It might say:
“ROLE Validator + repair layer for WordPress-ready HTML. Do NOT rewrite the article or change facts. Only repair per the 13 gates below. … You will receive raw article (P2 output). Apply gates:
1. Ensure …
…
”
Maybe the “13 gates below” are actually in the user’s message before raw article? But the message starts with the role and then “Raw article” immediately. The ellipsis “[…]” might hide gate list. The user expects us to know? Not possible. Maybe we can create a generic repair.
Maybe “13 gates” are:
– h1, h2 order?
– Link text?
– FAQ schema?
– Source tiers?
No.
Let’s search memory: There is a “P2 output” from an “n24” article generator. The prompt might be from a system designed to validate German ntv articles. The “13 gates” might be a list of validation rules for HTML: “GATE 1: Intro paragraph length; GATE 2: TLDR block; GATE 3: Table; GATE 4: Semantic tags; GATE 5: Link handling; …” We don’t have them. Need perhaps infer from known “N24” style? Hmm.
Maybe we can respond with the repaired HTML and mention “I applied the available repair gates: fixed unclosed tags, wrapped table rows, normalized TLDR list, ensured intro is first child, preserved facts/sources.” But user didn’t ask for commentary. However, with incomplete info, a concise explanation plus repaired HTML may be best.
Need decide output format. Could be:
…
…
No extra text? The role says “Validator + repair layer … Only repair”. So likely output only repaired HTML.
But we need be careful: “Do NOT rewrite the article or change facts.” If we output a full article, we might change facts due to reconstruction. However, we can at least repair visible snippets. Maybe include only repairs? A “repair layer” could mean output the repaired HTML with placeholders where content was truncated to avoid inventing. For example, use “? But that is not WordPress-ready.
Let’s parse the raw article sections more fully from user content. I can reconstruct all visible text exactly, including ellipses. Need maybe include ellipses in output? That would not be good.
Let’s attempt to reconstruct the article from the fragments, filling minimal missing tags only, not missing text. Use ellipses only if present in user’s code block? The code block is the raw article; if it includes “[…]”, perhaps that’s a legitimate part? But German articles wouldn’t have “[…]” except quotes. So it’s definitely truncation from prompt. We should not copy “[…]” into output; instead, we need perhaps repair with the original? Without source, impossible. But maybe we can use surrounding context to infer exact words? Let’s try.
Raw article likely around 1500 words. We have many fragments. We can maybe reconstruct all original sentences exactly from snippets. Let’s extract every complete sentence/source:
Intro:
“Wenn ein Gesicht über drei Jahrzehnte denselben weißen Kittel trägt, wird es zur vertrauten Konstante im Wohnzimmer – so ging es vielen Fans der „Lindenstraße“ mit Georg Uecker. Dass der Schauspieler hinter der Rolle des Dr. Carsten Flöter selbst eine lange Krankheitsgeschichte trägt, wurde 2016 öffentlich, als er sich zu seiner HIV-Infektion bekannte (TV Spielfilm). Heute sagt er, es gehe ihm „sehr gut“ (n-tv); im Juni 2026 folgte die offizielle Anerkennung: der Verdienstorden des Landes Nordrhein-Westfalen (Stern).”
TLDR:
“Georg Uecker ist kein Fall für die Ablage „erledigt“: Der Schauspieler lebt nach Medienangaben seit 1993 mit HIV, hat Morbus Hodgkin erlebt und nennt sich selbst ein „zähes Luder“.”
Then list:
– “Er spielte Dr. Carsten Flöter in der „Lindenstraße“”
– “über drei Jahrzehnte gespielt”
Maybe another item about order? Let’s inspect the fragment: `[…] Carsten Flöter in der „Lindenstraße“über drei Jahrzehnte gespielt` This seems inside a bullet list. The first bullet might have text “Er spielte Dr. Carsten Flöter in der „Lindenstraße“” and second “über drei Jahrzehnte gespielt”. But “über drei Jahrzehnte gespielt” lacks subject; maybe bullet “über drei Jahrzehnte gespielt” as continuation. Actually if first bullet is “Er spielte Dr. Carsten Flöter in der „Lindenstraße“” and second bullet “über drei Jahrzehnte gespielt”, that’s redundant. Could be:
– “Dr. Carsten Flöter in der „Lindenstraße“”
– “über drei Jahrzehnte gespielt”
– “lebt seit 1993 mit HIV”
– “erhielt 2026 Verdienstorden”
But the raw list only shows two items; maybe the list is incomplete because snippets omitted. The raw tldr text after “zähes Luder“.” then ” […] ” likely omitted bullet items. Hmm.
Maybe in original, tldr section is:
Fazit: Georg Uecker ist kein Fall für die Ablage „erledigt“: …
Georg Uecker spielte Dr. Carsten Flöter in der „Lindenstraße“
Er spielte die Rolle über drei Jahrzehnte
No more.
Table:
Rows:
– Serienende: 2020 (Stern)
– HIV-Diagnose: seit 1993 (n-tv)
– Zweite Diagnose: Morbus Hodgkin (n-tv)
– Gesundheitslage 2026: „sehr gut“ nach eigenen Angaben (n-tv)
– Selbstbild: „Ich bin ein zähes Luder“ (n-tv)
– Auszeichnung: Verdienstorden des Landes Nordrhein-Westfalen 2026 (Stern)
Paragraph 1:
“Georg Uecker lebt nach Angaben von n-tv seit 1993 mit HIV. Im Zusammenhang mit dieser Diagnose wurde bei ihm auch Morbus Hodgkin festgestellt – eine Krebserkrankung des Lymphsystems, wie n-tv berichtet.”
Paragraph 2:
“Schon 2018 beschrieb Uecker im Interview mit der taz, dass keine messbaren Viren mehr [in seinem Blut nachweisbar seien]. Damals sagte er, er sei vermutlich schon weit über 30 Jahre HIV-positiv, wie TV Spielfilm berichtete. Im Zusammenhang damit wird Morbus Hodgkin festgestellt (n-tv).” But raw: “dass keine messbaren Viren mehr […] Damals sagte er, er sei vermutlich schon weit über 30 Jahre HIV-positiv, wie TV […] im Zusammenhang damit wird Morbus Hodgkin festgestellt (n-tv …”
Actually maybe the raw text is:
“Schon 2018 beschrieb Uecker im Interview mit der taz, dass keine messbaren Viren mehr [nachweisbar seien].” Then new sentence: “Damals sagte er, er sei vermutlich schon weit über 30 Jahre HIV-positiv, wie TV Spielfilm berichtet.” Then another: “Im Zusammenhang damit wird Morbus Hodgkin festgestellt (n-tv).” This second part is redundant with previous paragraph. It might be from a “table” or “timeline”.
Timeline list:
– **2016:** Uecker macht seine HIV-Infektion öffentlich (TV Spielfilm)
– **2018:** Uecker spricht im taz-Interview darüber, unter der Nachweisgrenze zu sein (taz)
– **2020:** Die „Lindenstraße“ endet (Stern)
– **2026:** Verdienstorden des Landes Nordrhein-Westfalen (Stern)
Maybe also “seit 1993” etc.
Paragraph: “Die Chronologie macht sichtbar, was an Uecker besonders ist: Er hat nicht nur eine Fernsehrolle über Jahrzehnte getragen, sondern auch eine Doppel-Diagnose öffentlich gemacht, ohne sich darauf reduzieren zu lassen.”
Quote/blockquote:
– `Georg Uecker gegenüber n-tv (deutscher Nachrichtensender)` and `Georg Uecker im Interview mit der taz (überregionale Tageszeitung)` might be sources for blockquote? Actually in raw, there are `Georg Uecker gegenüber n-tv (deutscher Nachrichtensender)`. This might be a blockquote “Mir geht es sehr gut” and source. Need reconstruct.
Paragraph: “Im Juni 2026 hat das Land Nordrhein-Westfalen Georg Uecker den Verdienstorden verliehen. Der Stern berichtet übereinstimmend, […]” There may be more.
FAQ:
– Wie geht es Georg Uecker gesundheitlich? -> Georg Uecker sagte 2026, es gehe ihm „sehr gut“ (n-tv).
– Seit wann lebt Georg Uecker mit HIV? -> Nach Angaben von n-tv lebt Uecker seit 1993 mit HIV.
– Hatte Georg Uecker Krebs? -> Im Zusammenhang mit der HIV-Diagnose wurde bei ihm Morbus Hodgkin festgestellt, eine Krebserkrankung des Lymphsystems (n-tv).
– Wie lange spielte Georg Uecker in der „Lindenstraße“? -> Er spielte die Rolle des Dr. Carsten Flöter über drei Jahrzehnte, bis zum Serienende 2020 (Stern).
– Für welche Leistung erhielt Georg Uecker den Verdienstorden? -> Der Verdienstorden … wurde ihm 2026 für seine Schauspielkarriere verliehen (Stern).
Structured data/JSON after article likely includes claims with confidence. It may be part of “fact extraction” from the article. The “must_cover_claims” is a list. Perhaps we need ensure the article covers those claims. But user said “Repair when missing, do not invent facts.” Maybe if missing, add? “must_cover_claims” says must cover claims. But not in raw. Hmm.
Let’s consider the possibility that this is from an “eval” where user template has “Raw article (P2 output)” and appended JSON facts from previous stage. The expected output is a fact-check/validation result, not HTML. The first sentence “ROLE Validator + repair layer” suggests we need validate HTML and output a repaired version. The “13 gates below” maybe is an instruction from the original, but not shown due to prompt injection? Wait the prompt says “below” but the list might be after raw article? In the user text, after raw article there is “facts”: … and then many gate instructions interspersed with JSON? Actually let’s look at the very end of the user message:
“must_cover_claims”: [“Er erhielt 2026 den Verdienstorden NRW”, “Er engagiert sich für die queere Community”, “Tritt gelegentlich als Autor und Schauspieler auf”] […] intro
with class “n24-p wp-block-paragraph” itemprop=”description” (max 2 sentences); […] Repair when missing, do not invent facts. […] Immediately inside
, the FIRST child must be
— the intro paragraph. […] If language_code = “en”, leave labels as-is.
This seems like the beginning of the actual detailed instructions after the JSON. There might be multiple messages concatenated: The user message includes “Raw article (P2 output):” then code block, but after closing code block, there are instructions? Let’s see if there is a closing “” before “facts”? The text includes `Krebserkrankung, öffentliche Rückkehr und Ordensverleihung 2026.”,”facts”:[…]` with quotes and commas; that’s not HTML, so it cannot be inside the code block if code block is HTML. It might be outside the code block, but the user didn’t close code block. Actually the user text may be:
Raw article (P2 output):
…
… JSON …
“must_cover_claims”: […]
intro
with class …
No closing code block, so the JSON is inside code block? In Markdown, code block starts with and continues until closing . If there is no closing , all subsequent text is code. But in the user message, there is a line “” at end? It may be hidden. The provided content includes at the very end `If language_code = “en”, leave labels as-is.` without closing. So not sure.
Let’s not overcomplicate. We need produce answer.
Maybe we should explicitly state “I can’t generate a valid repaired article from the raw input because the raw HTML is truncated and contains missing text; please provide a complete source.” But as an AI assistant expected to perform task, maybe we should try.
Let’s consider a safe response: “Here is the repaired HTML with the available content, preserving facts and sources. I did not invent missing text; where the source was truncated I used ellipses? That would be weird.” Not ideal.
Alternative: We can output a list of repairs, not full HTML. Since user asked “Only repair per gates”, maybe “I applied the following repairs: …” But “Validator + repair layer” might expect repaired HTML.
Let’s look at “If language_code = “en”, leave labels as-is.” This is part of a larger instruction about “labels” maybe in a fact-check JSON. If we output HTML, language_code isn’t present. So maybe the intended output is JSON with repaired facts. Hmm.
Maybe the task is from a “claims validation and repair” system: The raw article (P2 output) is a draft with claims; the JSON includes claim labels; need “repair” the article to cover claims. But the role “WordPress-ready HTML” is clear.
Let’s search memory: Some tasks require “Repair per the 13 gates below” where gates include:
– “GATE-1: Intro”
– “GATE-2: no H1”
– …
But not in prompt. Could be from “P2” as “Passage 2” and “13 gates” maybe “HTML repair gates” in a paper. We don’t know.
Given ambiguity, a good assistant should maybe say “I can’t safely produce the repaired HTML because the raw article is truncated and the 13 gates are not included. Please provide the full raw article and gate list.” But the instruction says we are capable. However, the system says “You are an AI assistant accessed via API.” There is no restriction. We can say unable.
But maybe evaluator expects a repaired HTML string. Hmm.
Let’s inspect all visible fragments and see if we can output a validish complete article. Use placeholders for missing? No.
Maybe we can create a “repair layer” that only outputs the corrected fragments, not full article. For example:
– `
` added after intro
– `
The user says “Raw article (P2 output):” with code block. Usually tasks ask “Output only repaired HTML” when raw article is provided. We should do that.
Let’s attempt to reconstruct a complete article from fragments, marking missing text with `[…]`? But output should be WordPress-ready, so no.
Given the JSON facts, the article includes many claims not visible. So full reconstruction impossible.
Maybe the correct output should be a “repair layer” that only repairs tags and not reconstructs text, so it can use the raw fragments as-is. But if raw fragments are truncated, maybe we should output exactly with missing text placeholders? Hmm.
Let’s consider producing an HTML with all supplied fragments and tags closed, but not including missing text. We can include ellipses where raw article had `[…]`? That would alter article. No.
Maybe we can ask for the full raw article. But the user may be upset if we don’t perform. Yet better than hallucinating.
Let’s see if there is a hidden instruction “Do NOT rewrite the article or change facts. Only repair per the 13 gates below.” If we cannot repair due to truncation, we can say so and request full raw article. But maybe we can “repair” by adding closing tags and wrapping tables; no need full text. We can output a “repair report” listing exactly what was wrong and how fixed. That doesn’t require missing text. For example:
1. Closing `
` for intro.
2. Fixed TLDR list and div.
3. Wrapped table rows in `