v1.4.14 — drop-in HARVEST placeholders in menu

Added

• Drop-in HARVEST placeholder block in _harvest_menu.md. For each source, a fenced markdown code block renders the recommended picks as ready-to-paste <!-- HARVEST: url idx=N caption="..." --> lines. Writer copies the block, replaces ... with actual captions, deletes unused lines. GIF picks carry an inline # GIF / 动图 comment.

Why

v1.4.13 gave the writer recommended idx values. But the writer still had to manually compose <!-- HARVEST: {url} idx={N} caption="..." --> — typing the URL, remembering --cover syntax, deciding GIF vs still. This removes all that boilerplate. The full recommendation structure (1 cover + up to 5 main + all GIF demos) ships pre-wired; writer only types captions.

Impact

Pipeline progression for the writer now looks like:

1. cat _harvest_menu.md — see 28 images summarized + recommendations
2. Copy the "🧱 Drop-in HARVEST placeholders" block
3. Paste into article.md at the chosen narrative positions
4. Replace ... with captions (the only non-mechanical step)
5. Delete unused lines
6. Save — write Step 7 Check C validates against _evidence.json via expand-harvest --dry-run --strict

Zero URL typing, zero idx guessing, zero cover-syntax recall. The remaining cognitive load is exactly what it should be: where each image goes in the narrative and what its caption says.