NeuroVim

Changed

• The CIPHER uplink now comes from @neurovim/core. Prompt building, the chat session and the turn choreography moved upstream behind the new LlmPort; this plugin keeps what is genuinely its own — transport, SSE, endpoint resolution and model choice — and wires them into the port via CorePortAdapter. Behaviour is unchanged, including the distinction between CUT (the killed turn still lands its partial answer as "— signal cut") and RST (the turn is disowned and the channel cleared). 489 lines removed, 48 added.

Added

• THE ARCHIVE — new hub tab browsing the story layer: 9 loot artifacts and 14 intercepted fragments, read in a modal with rendered markdown. Loot unlocks with your level; fragments are open from the start. The unlock state was already being tracked — this makes it visible.
• Archive groups collapse, and their headers show how many artifacts of the group are readable (1/9) — so a long loot list no longer buries the fragments below it. The collapsed state is remembered.

Fixed

• Hub tabs wrap to a second row instead of running off the pane edge. In a narrow sidebar the last tab (UPLINK) was unreachable.
• The lore reader now follows the CRT colour scheme. It was built as a twin of the briefing modal but never inherited its palette rules, so artifacts rendered in the theme's colours while the hub behind them was green.
• ASCII headers in the reader scroll instead of wrapping — wrapping broke every box frame apart.
• Archive cards align consistently. Unlocked cards are <button>s and locked ones <div>s; Obsidian centres button content, so the two variants disagreed.