Direction lock: Lehi sprint plan, revised Codex handoff, canonical 8/9 handoff - #6

…onical handoff

Applies Brandon's confirmed decisions: wedge->ladder->referral flywheel;
door demo is subscription ChatGPT on the prospect's phone; the paid API
probe becomes a deferred internal proof lane (no key, no paid calls this
phase); territory Lehi, Utah. Revises the build plan and Codex handoff
accordingly (probe calibration no longer the launch gate; skills install
deduplicated to the canonical ~/Projects/skills store), adds the Lehi
20-door sprint plan with the door-demo procedure and pass/fail evidence,
fixes the one honesty violation on the price card, records probe state
(main 450f25d, gpt-5.6-terra, 8/8 tests), and issues the 2026-08-09
canonical handoff superseding 8/7.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

Updates stale probe receipts across the canonical handoff, Codex work
order, build plan, and direction-lock journal: main now
0e455bfd200da0d3f64892c6f6915de9cb77d882 with sprinkler, garage_door,
auto_repair added to questions.yaml; 8/8 tests, import and secret scan
clean, repo private. Codex launch lane marked complete with no remaining
task; skills and paid API lane stay deferred, unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

…nal spec

Brandon verified no Guild GHL account exists; the only HighLevel access
on the Mac is a location-scoped client sub-account integration that is
off-limits for Guild use. Removes the GHL login/funnel/number launch
prerequisite as a false assumption. Adds the offline Lehi field log
(self-contained HTML, localStorage autosave, progress /20, escaped CSV
export, typed-RESET-gated erase, no network calls; 40/40 Playwright
behavioral checks) plus a CSV header template as the canonical record
for the first 20 doors. Sprint plan, build plan, canonical handoff,
journal, and funnel build sheet reconciled: funnel, dedicated number,
and card printing are post-signal procurement after a close; prices,
doctrine, cohorts, demo wording, and evidence gates unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

Doors enforced 1-20 and unique (duplicates rejected, progress counts
unique doors, saves lock behind a sprint-complete state after 20).
Price cohort is derived from door parity (odd $497 / even $997),
displayed read-only, never manually selectable; price_cohort export
preserved. Demo consistency enforced by UI locks (no -> result locked
to not-run; yes -> not-run unselectable, present/absent required;
switching clears incompatible state) plus a second guard in submit
validation; novalidate added so the JS guards give uniform rejection
messages. CSV export now door-sorted. Behavioral suite extended and
green at 71/71, covering duplicate/out-of-range/21st-entry rejection,
completion lock persistence, cohort derivation in UI and CSV, and both
invalid demo combinations. Receipts updated in the canonical handoff
and direction-lock journal.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

One self-contained file with five in-page views: COMMAND (mission,
phases, zero-cost readiness), WORKFLOW (door flow with locked honest
wording + fulfillment lanes tagged by tier), LOG (the full field log on
the same localStorage keys), CLIENT (prospect-facing presentation mode
with editable prompt, honest present/absent reads, and the current
door's locked price only), RESULTS (live stats with the correct
demo-rate denominator and a decision gate that never labels a partial
run). Resolves the measurement schema mismatch by adding owner_present
and sit_down_booked with enforced truth cascades and non-destructive
legacy migration; CSV header and template updated. Committed behavioral
suite (playbooks/tests/command-center-smoke.js) passes 126/126 at 390px
and desktop, including zero-external-request and guarantee-language
checks. Sprint plan, build plan, canonical handoff, and journal now
point to the Command Center as the canonical first-sprint surface; the
standalone field log is demoted to emergency fallback.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

…meframe

CLIENT price now derives from the LOG's active door (a valid unlogged
1-20 in the door input) with nextDoor() fallback and TARGET at
completion, so out-of-sequence doors present the correct locked price.
Close copy corrected from 'within days' to the approved 'within the
week'. Committed regression checks added (active door 2 -> $997 and
door 5 -> $497 before any saves, plus copy assertions); suite green at
130/130 on mobile and desktop across two consecutive runs.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017SgKh5sTzx4rjanwEAXtfc