paper-glosslisted

# Paper Gloss — click-to-reveal jargon glosses A companion to `/paper-eli5`. Takes a finished eli5 output and produces a second pass where every approved technical term becomes **clickable at each occurrence** — tapping it pops up a plain-English expansion, then dismisses. The original prose is never rewritten or interrupted; the paper reads exactly as it did, with jargon quietly wired up for lookup instead of permanently annotated inline. Example: "The model is trained using **backpropagation**" — the term `backpropagation` gets a dotted underline. Clicking it opens a small popup: *"a step where errors are traced backward through each layer to adjust the model's internal dials."* Clicking again, clicking elsewhere, or `Escape` closes it. The same expansion also appears as a row in the Glossary panel (📖, fixed top-right) so every expansion is still visible at a glance without touching the running text. Because this is genuinely interactive, the deliverable is a self-contained HTML page, not Markdown — delivered both as a file in the repo and as a published claude.ai Artifact. --- ## Parse `$ARGUMENTS` - **Local path** to a `-eli5.md` file → the target document. - **`--retrofit <glossed.html> [artifact-url]`** → skip Phases 1–2 and run RETROFIT mode (below): typeset math in an already-published page and redeploy it to the same Artifact URL. - **`--annotate <glossed.html> [artifact-url]`** → skip Phases 1–2 and run ANNOTATE mode (below): inject the annotation layer