herdr: native-Windows launch surface (agent.start launch, pwsh parked dialect, Windows-host re-characterization)

[pbean]

The herdr backend (#136 run path, #137 TUI-launch surface) ships end-to-end on POSIX. Its native-Windows launch surface was deliberately deferred ("PR-6" in the PR bodies) and currently has no open tracker — the umbrella #92 is closed. This issue is that residue.

Remaining work

1. win32 launch path via herdr agent.start — the POSIX launch types exec <argv> into a fresh tab so process-exit == pane-close == window death (herdr_backend._launch); exec has no Windows equivalent, and agent.start (which splits the active tab) is the reserved win32 launch verb. Source: herdr_backend.py module-docstring degradation ledger + the _launch comment.
2. pwsh parked-window dialect — new_parked_window types a POSIX exec sh -c '<argv>; ec=$?; banner; read -r; trailer' recipe (the trailer cats the per-window return file). The pwsh dialect of the recipe + trailer is deferred alongside the launch path. Source: degradation ledger ("POSIX-only, like the exec launch (pwsh dialect is the Windows follow-up)").
3. Re-run Phase-0 characterization on a real Windows host — the shipped backend is POSIX-verified / Windows-by-construction; no Windows box was available during feat(mux): herdr cross-platform multiplexer backend #136/feat(mux): herdr TUI-launch surface #137. Per docs/ROADMAP.md: "the Phase-0 characterization must be re-run on a real Windows host before Feature Request: Support for Native Windows (Remove tmux dependency for Windows developers) #92 is claimed."

Non-goals here

• The socket transport replacing subprocess _HerdrClient is a separate optional enhancement (the class boundary already isolates it).
• The sibling tmux-family win32 backends (psmux re-land: Native-Windows psmux multiplexer backend: re-land as a thin leaf over the dialect seams #58; tmux-windows: Add native Windows tmux backend #85) are tracked separately — herdr is pairwise-discriminating against them by construction (which("herdr")).

Selection note (no code change needed when this lands)

herdr registers matches=True but _PLATFORM_DEFAULTS still declares psmux the win32 default, so herdr activates on win32 only via first-match while no higher-priority backend is available — once psmux ships, the default reasserts automatically.

Refs: #92 (closed umbrella), #136, #137, #58, docs/ROADMAP.md "Native Windows multiplexer backend", docs/porting-to-a-new-os.md.

[coderabbitai]

🔗 Related PRs

#47 - refactor(adapters): extract the shell-dialect seam from new_window/new_parked_window [merged]
#72 - fix(install): make psutil a hard win32 dependency [merged]
#118 - Multiplexer selection: availability-aware defaults + dev-chosen backend (#87) [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[pbean]

The herdr backend moved out of core to pbean/bmad-loop-adapter-herdr (#146, discovery seam in #145), so the win32 launch surface goes with it — this is now tracked as bmad-loop-adapter-herdr#1. Core-side native-Windows work (the psmux backend, #85/#87 lineage) is unaffected; #143's TUI raiser-side guards also stay in core.