1. Extraction prompt: Use only the files in ./inputs. Read inputs/Classification Task.md and all .tex and .bib files under inputs.
Create outputs/theorem_map.md identifying the main proved theorems and the mathematical objects leading up to each theorem.
Follow the structure in inputs/Classification Task.md. Do not search the web. Do not invent theorem statements or proof dependencies.

2. Redaction prompt: Read inputs/Redaction Task.md and complete the task.

*Manually remove items 7 and 8 from the redacted seed files.


3. Generation prompt: Read inputs/Generation Task.md and complete the task.


4. Scoring: This isn't a prompt. Run these python files:
python eval_deepeval_quality.py
python eval_deepeval_alignment.py