refactor(daemon): split session-state handlers by device-state domain question

[thymikee]

Summary

Splits the two src/daemon/handlers/session-state.ts functions whose cyclomatic complexity alone kept CRAP ≥ 30 at 92–96% coverage (G2 of #3424, part of #3414):

• handleSessionStateCommands: 34 → 4 — now a dispatcher over dedicated boot / shutdown handlers.
• handleAppStateCommand: 30 → 5 — appstate separates target admission (session-appstate.ts), the Apple session response (session-appstate-apple-session.ts), and the device-runtime query (session-appstate-device-runtime.ts).

All resulting functions score below 30 (check:coverage-crap --all and fallow health at max-crap 1 show no production finding ≥ 30 in the new modules; highest is handleShutdownCommand at CC 15). New modules are registered in the R81 daemon-layer manifest; the path-keyed session-state.ts health-baseline entry moves in the final chore(gates) commit (no bulk regeneration).

Behavior is shown unchanged: existing tests were kept green with moves only (renames + one source-slice test retargeted to session-shutdown.ts, one inventory-mocked boot test adapted to the boot file's harness), no assertion edits. Production gross diff is 975 lines (6 files).

Part of #3424.

Validation

• Tested SHA: ab79e9b8f
• pnpm check:affected --run: all runnable checks passed (under the machine lock).
• pnpm check:coverage-crap --all: "no function(s) at or above 30" (partial map including the six handler modules).
• pnpm check:layering, pnpm check:fallow, pnpm typecheck: green.

[github-actions]

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	5.17 MB	5.17 MB	+568 B
Package (unpacked)	5.17 MB	5.17 MB	+568 B
Package (download)	1.55 MB	1.55 MB	+156 B

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	30.9 ms	29.1 ms	-1.7 ms
CLI --help	83.5 ms	84.2 ms	+0.7 ms

[cubic-dev-ai]

All reported issues were addressed across 16 files

<sub>Reply with feedback, questions, or to request a fix.

View guided diff | Turn on auto-fix | Re-trigger cubic</sub>

[thymikee]

The split of the session-state handlers by domain question looks correct at ab79e9b, and all 19 checks pass. I did not run the test suite, check:coverage-crap or fallow myself, so the CRAP<30 claim rests on your report and CI. No conflicts.

Not blocking, and you can take or leave these: (1) the full diff is +1214/-1023 (2237 gross), mostly test moves, while docs/agents/pull-requests.md and #3424 set a 1,000-line gross budget and the rename-only exemption does not apply, so either move the tests first in a refactor(move) PR (session-boot-shutdown.test.ts to session-boot.test.ts and session-shutdown.test.ts) or state the overage against the gross number in the PR body; (2) the routing test at https://github.com/callstack/agent-device/blob/ab79e9b/src/daemon/handlers/__tests__/session-state.test.ts#L26 only asserts INVALID_ARGS, which all three commands return, so swapping boot and shutdown routing would still pass, and asserting the command name in the message would fix that; (3) the Android serial allowlist sort is now spelled in session-boot.ts:119 and session-appstate-device-runtime.ts:15, and also in session-doctor.ts:276 and session-lifecycle/internal/inventory.ts:373, so one shared helper beside resolveAndroidSerialAllowlist in @agent-device/kernel/device-isolation could serve all four sites.

Could session.ts's existing command table point boot, shutdown and appstate straight at handleBootCommand, handleShutdownCommand and handleAppStateCommand, so session-state.ts (https://github.com/callstack/agent-device/blob/ab79e9b/src/daemon/handlers/session-state.ts#L12) and its routing test can go, since the table is compile-checked and the new switch has a return null branch it cannot reach? Could session-appstate-device-runtime.ts also fold back into session-appstate.ts, since that is its only caller?

The cubic-dev-ai test-assertion thread at #3450 (comment) is fixed at this head, so you can resolve it.

Before merge, please answer the gross-diff budget (2237 vs 1,000) and whether session.ts should own the dispatch. Also, #3424 is labeled on-hold pending coordination with the src/daemon to packages/daemon extraction, and I did not check whether that move collides with these new files, so please confirm that coordination before landing new handler files.