Second pass over the 47 lower-confidence orphan candidates held back
from the earlier 34-file cleanup. This time checked two signals, not
one: (1) is the script's own filename referenced anywhere, and (2)
does the script's own Temp/*.json output get read by any other file
(a script can be "orphaned" by name but still load-bearing if
something else consumes what it produces).
7 of the 47 failed check (2) - their outputs (outcome_ledger_v1.json,
pre_distribution_early_warning_v3.json, shadow_ledger_v1.json,
calibration_registry_v1.json, final_execution_decision_v4.json,
live_outcome_ledger_v1.json, final_decision_packet_v2.json) are read
by other tools even though the producing script's name never appears
elsewhere - kept those 7 in place.
The remaining 40 (WBS-ticket-tagged one-offs: wbs81/92/93/95/96_*,
build_p0-p6_*; and misc build_*/validate_* diagnostics with no
external reference by name or by output) were deleted. Re-ran the
full-repo reference search (zero hits) and quick-mode release DAG +
WBS validator afterward - failure set is unchanged (same 10
tasks already known blocked on missing live trading data / a running
web app + Playwright, unrelated to this deletion).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
kjh2064
73572b6211