paper-figureslisted

# Paper Figures Run the bundled extraction script with the interpreter and paths rendered by the repository installer: ```bash "{{PAPER_FIGURES_PYTHON}}" "{{PAPER_FIGURES_SCRIPT}}" \ "<paper.pdf>" "<output-directory>" [options] ``` Before running it, verify that `{{PAPER_FIGURES_PYTHON}}` and `{{PAPER_FIGURES_SCRIPT}}` were rendered and that both paths exist. If either placeholder remains literal, stop and tell the user to rerun the repository installer. For an `ai-wiki` paper note, use: ```bash "{{PAPER_FIGURES_PYTHON}}" "{{PAPER_FIGURES_SCRIPT}}" \ "<paper.pdf>" \ "{{AI_WIKI_DIR}}/_attachments/paper-figures/<paper-slug>" ``` `{{AI_WIKI_DIR}}` is installer-rendered and defaults to `0ai_wiki`. The paper slug must match the paper-note filename without `.md`. ## Options - `--zoom 2.0` — render zoom; 2.0 is approximately 144 DPI and is the default. - `--pages 1,4,9` — additionally render selected 1-indexed pages as `page_N_full.png`. - `--figures-only` — extract figure captions only. - `--tables-only` — extract table captions only. - `--max-fig-height-ratio 0.85` — cap the inferred crop height as a fraction of page height. `--figures-only` and `--tables-only` are mutually exclusive. ## Output contract The script writes: - `figN.png` for each detected figure caption. - `tableN.png` for each detected table caption. - `page_N_full.png` for requested full pages. - A JSON manifest on stdout with `pdf`, `out_dir`, `figures`, `tables`, and `full_pages`. Each asset ent