mplex: wiring doc notes

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

taicpp/docs/wiring/mplex.md

@@ -39,6 +39,14 @@ if (opts.mplex.has_value()) {
 like the Python (`sessions should be nested with care, unset TAI_PLEXER_ACTIVE to force`,
 `open terminal failed: not a terminal`, `no server running on <sock>`, ...).
 
+Argument quirk to keep in step with tai.py: argparse's `nargs=REMAINDER` swallows a literal `--`
+inside the mplex words (tai.py: `--mplex send-keys -X -- cursor-up` loses the `--`), while
+`tests/mplex_cli_main.cpp` passes argv through untouched. The `src/cli` parser must drop/keep `--`
+exactly as `argparse.REMAINDER` does (verify against `python3 tai.py --mplex ls -- x`).
+
+The `version` format variable is `detz::kVersion` when the build puts the generated `version.hpp`
+on the include path (the Makefile's `-I$(GEN_INC)` does), `1.44.2` otherwise.
+
 ## `--plexer-wrap`
 
 tai.py 40129-40133: same `home` computation, `profile = args.profile or "default"`:
@@ -79,6 +87,6 @@ behave as in tai.py.
 ## Verified
 
 See the report of the porting run: `tests/mplex_flow_test.cpp` (spec/flow corpus against tai.py),
-`tests/mplex_e2e.py` (six scenarios run through a pty against both `tai.py --mplex` and
+`tests/mplex_e2e.py` (seven scenarios run through a pty against both `tai.py --mplex` and
 `build/mplex_cli --mplex`, observations compared), `tests/mplex_tmux_oracle.py` (tmux 3.6 oracle for
 the tmux controller).