Add glossary decks

CLAUDE.md

@@ -48,6 +48,12 @@ data/ ──[Stage 2: amateurfunk_anki.py]──► anki/
                                             ├── amateurfunk-technische-kenntnisse-a.apkg
                                             ├── amateurfunk-betriebliche-kenntnisse.apkg
                                             └── amateurfunk-kenntnisse-von-vorschriften.apkg
+
+shorthand.json ──[Stage 2b: amateurfunk_shorthand.py]──► anki/
+                                            └── amateurfunk-abkuerzungen-q-gruppen.apkg
+
+technical.json ──[Stage 2c: amateurfunk_technical.py]──► anki/
+                                            └── amateurfunk-technische-abkuerzungen.apkg
 ```
 
 ### Stage 1 — `amateurfunk_fetch.py`
@@ -92,6 +98,45 @@ data/ ──[Stage 2: amateurfunk_anki.py]──► anki/
 The Anki design decisions (shuffle seeding, deterministic build epoch,
 SVG dark-mode handling, schema choices) live in `DESIGN.md` §7.
 
+### Stage 2b — `amateurfunk_shorthand.py`
+
+A sibling builder for a standalone reference deck of Q-groups and
+operating abbreviations — the ones in the exam plus the most common
+on-air shorthand the exam never covers (real operating knowledge, not
+just the test). Content lives in the hand-curated `shorthand.json`
+(editorial, tracked in git, like `explanations.json`); the
+`references/Q-Codes.md` reference is where the exam-present codes were
+catalogued.
+
+Each code is one Anki *note* with two card templates — forward
+(code → meaning) and reverse (meaning → code) — so a single record
+drives both directions. A Q-group means one thing as a statement
+(`QSO`) and another as a question (`QSO?`), so each Q-group yields two
+notes; plain abbreviations yield one. All IDs/GUIDs are hashed from the
+displayed code form (stable re-import). The deck is catalog-independent
+and fully deterministic; it only consults `data/` to borrow the
+manifest build epoch when present. Low-level apkg/SQLite machinery is
+imported from `amateurfunk_anki.py` so the two stay in lockstep. The
+glossary machinery shared with Stage 2c (two-template note type,
+two-cards-per-note writer, packager, build-epoch resolver, entry
+validator) also lives here.
+
+### Stage 2c — `amateurfunk_technical.py`
+
+A third glossary deck, same card mechanics as Stage 2b (one note,
+forward+reverse templates), for the *technical* vocabulary rather than
+operating shorthand: modulation/modes (SSB, FM, CW), signal domains
+(NF, HF, ZF), building blocks (VFO, PLL, AGC), components, measurements
+(dB, SWR, PEP), propagation, digital modes, and the
+organisations/regulations (ITU, CEPT, EMV) — exam terms plus common HAM
+abbreviations beyond the exam. Content lives in the curated
+`technical.json`; each entry carries a German `category` (Betriebsart,
+Bauteil, …) shown on the card and used as a `kategorie-*` tag. The
+shared glossary machinery is imported from `amateurfunk_shorthand.py`;
+this script only adds the data shape, the deck/model names, and the tag
+scheme. IDs live in their own `technical` namespace so the two glossary
+decks never collide on import.
+
 ## Repo conventions
 
 - Python 3.11+, standard library only. No third-party dependencies