[Bug] H1: Missing screen/handler functions cause infinite error loop in app runtime

[fissible]

Source: external review report, 2026-08-24 (severity: High)

Problem

Screens in the v2 app runtime are string-driven with no existence checks:

• src/app.sh:115 — _type=$("${_prefix}_${_current}_type")
• src/app.sh:158 — "${_prefix}_${_current}_${_event}"

A typo'd screen name, a handler that forgets to set _SHELLFRAME_APP_NEXT, or an unmapped event yields _current="". The next iteration then invokes PREFIX__type → command-not-found → empty type → case falls through → _rc=0 → nonexistent PREFIX__ handler → loops forever spewing errors. Only SIGINT gets you out.

What it needs

• declare -F existence check on the _type function and on the event handler before invoking
• Non-empty assertion on _SHELLFRAME_APP_NEXT after a handler returns
• A clear diagnostic (shellframe: unknown screen 'foo' (prefix 'bar')) and a non-zero exit instead of a loop

What it delivers

First-five-minutes robustness for anyone adopting the app runtime — a typo becomes a readable error instead of a runaway process.

Tests

• ptyunit case: app with a handler that never sets NEXT → exits non-zero with diagnostic, no loop
• ptyunit case: transition to a screen with no _type function → same

Effort: S
Deps: none
Priority: pre-OSS-launch blocker (one of #41/#42/#43/#44)

[fissible]

Fixed: declare -F guards on the screen type fn + event handler, non-empty NEXT assertion, and a shellframe: unknown screen ... stderr diagnostic with exit 1. Covered by 4 new watchdog-protected unit tests in tests/unit/test-app.sh (22/22 file, 1519/1519 suite).