I have uploaded ATR prompt files as separate .txt files or as a small prompt-pack ZIP. Please do not summarize the prompt files. Use them as operating instructions.

First, list the uploaded prompt filenames you can access. Then identify which ATR prompts are relevant to this task and which are missing, if any.

Task type: [HTML/Blogger translation | quick raw text to AtR Blogger HTML | clean translation | review only | Tibetan/Indic source verification | East Asian source restoration | Baihua | prompt-suite maintenance]
Target language: [Simplified Chinese / Traditional Chinese / English / other]
Source file or source text: [uploaded file name or pasted below]
Output needed: [clean text / Blogger HTML / DOCX / QA report / repaired artifact / other]

Required workflow:
1. If this is quick raw text to AtR Blogger HTML, use the Quick Blogger Formatting Prompt first and do not translate, polish, summarize, or omit any wording. If this is HTML/Blogger translation or complex Blogger work, use Prompt A as the controller and preserve all HTML structure, links, wrappers, metadata, and safe prompt-body escaping.
2. If this is a clean translation, draft with Prompt 1 unless another uploaded prompt is more specific.
3. If Tibetan, Sanskrit, Pāli, Prakrit, or Indic/canonical source verification is involved, activate Prompt T before relying on any public English translation. Public translations are witnesses only, not authorities.
4. If Chinese, Japanese, Korean, Chan/Zen, Dōgen, CBETA, SAT, or East Asian source restoration is involved, activate Prompt X. Restored source blocks must not replace complete body translation coverage.
5. If classical Chinese needs modern Chinese explanation, use Prompt 5, but do not add unsupported Dharma commentary into the 白话.
6. After drafting or repairing, run Prompt 6 as the high-fidelity review gate.
7. Use Prompt 9 only after Prompt 6 when whole-document refinement, consistency, or publication polish is needed.
8. For HTML/Blogger artifacts, run the Strict HTML QA Audit Prompt before claiming the file is Blogger-ready.
9. If creating or editing files, read back the saved artifact before claiming completion.
10. If any required prompt file is missing or unreadable, say exactly which file is missing and continue with the safest partial workflow rather than inventing prompt content.

Now proceed with the task using the uploaded ATR prompt files.

Basis/Dharmakaya QA6 controller addendum — old/new styled source and restored-block workflow:
11. If uploaded files include an old source, new styled source, target-language draft, prompt pack, and QA reports, first list files by role: Content authority, Styling shell, Existing target-language draft, Prompt pack files, and QA reports / older generated artifacts.
12. If the user says the new styled file should have the same content as the old file, run normalized visible-text content parity before translation. Repair the styled source first. Do not translate from a defective styled source, and do not use paragraph count alone as evidence.
13. If repaired source restores large missing blocks and a target-language artifact already exists, translate the restored blocks freshly from the source using Prompt 1. Do not rely on existing public or old translations unless the user explicitly permits it. Then run Prompt 6, Prompt 9 where appropriate, and Strict HTML QA.
14. If any restored block remains untranslated, label the artifact exactly as STRUCTURALLY READY BUT TRANSLATION INCOMPLETE. Do not say final, complete, or Blogger-ready.
15. For complex repair sessions, return final repaired source HTML, final target-language HTML, QA report TXT/JSON, optional ZIP containing outputs, and a separate prompt-patch ZIP if prompts need updates.
16. In the final response, state only what was actually verified. Do not claim all issues fixed unless the QA report supports it.

Dzogchen terminology controller addendum — gnas tshul / snang tshul, sentient-beings, cognizance, obscurations, and phenomena QA:
17. If the task involves Chinese Buddhist/Dzogchen translation, review, formatting, or artifact QA, activate the relevant terminology gate before final claims.
18. Search the exact target artifact for 知识 / 知識, 有情的显现样态 / 有情的顯現樣態, 安住样态 / 安住樣態, 诸现象 / 諸現象, 知性, 宇宙与有情 / 宇宙與有情, and 被显为清净 / 被顯為清淨.
19. For each hit, inspect source context and classify fixed / intentionally retained / needs source review. Do not apply these as free-polish substitutions.
20. Preserve gnas tshul / snang tshul distinction, avoid 有情的显现样态 where it means mode of appearance for sentient beings, prefer 器情世间 for snod bcud-style universe-and-beings formulas, avoid 知性 for Dzogchen cognizance, use 遮障 for doctrinal obscurations and 诸法 for doctrinal phenomena.
21. Do not claim final, complete, Blogger-ready, or terminology cleanup complete unless these target-side searches and source-context classifications were performed on the exact returned artifact.

Batch 33 controller addendum — dependent designation / effort / View / common-uncommon Mahāyāna gate:
22. If the task involves Chinese Buddhist/Dzogchen/Madhyamaka translation, review, formatting, or artifact QA, activate the Batch 33 terminology gate before final claims.
23. Search the exact target artifact for 依名假立, 依缘假立 / 依緣假立, 努力, 精进 / 精進, 勤作, 知见 / 知見, 共同大乘, 不共大乘, 非共同大乘, and 特殊大乘.
24. For each hit group, inspect source context and classify fixed / intentionally retained / needs source review. Do not apply these as free-polish substitutions.
25. Preserve dependent designation as 依缘/依蕴/依支分/假名安立 by context; preserve Dzogchen/Mahāmudrā effort as 勤作/no-effort wording where appropriate; preserve View/lta ba as 见/见地/见解 rather than automatic 知见; preserve common/uncommon Mahāyāna as 共同大乘/不共大乘.
26. Do not claim final, complete, Blogger-ready, or terminology cleanup complete unless these target-side searches and source-context classifications were performed on the exact returned artifact.
