Run git status. Expect: fatal: not a git repository (or any of the parent directories):
.git — byte-identical, no partial .git anywhere above this directory.

Run ls -la ~/Projects/krtv. Expect: the only product artifact at top level is
files.zip; everything else present (AGENTS.md, CLAUDE.md, _handoff/, .claude/,
think-like-fable/) is kit scaffolding, not krtv code — no src/, apps/, package.json,
or node_modules anywhere.

Run find ~/Projects/krtv -maxdepth 1 -iname "readme*". Expect: no output — no README
exists at the project root.

Final check: curl -s "http://127.0.0.1:8787/api/kanban?project=krtv" — krtv's only card in
the backlog column reads no root README found, and every other column (in_progress,
running, in_development, testing, done, deployed) is empty for this project. The board
and the disk agree: nothing has been built.

Run stat -f "%z bytes" ~/Projects/krtv/files.zip. Expect: 20830 bytes.

Run unzip -l ~/Projects/krtv/files.zip. Expect: exactly 5 entries —
PACKET.md (17755 bytes), WIREFRAMES.html (17126 bytes), NOTES.md (2635 bytes),
krtv-prod-ready-plan.md (10874 bytes), plan-queue.md (369 bytes) — totaling 48759.

Unzip to a scratch directory — never overwrite files.zip itself:
unzip ~/Projects/krtv/files.zip -d /tmp/krtv-check. Expect: all 5 files extract without a
CRC/error, and none is 0 bytes.

Final check: head -c 40 /tmp/krtv-check/PACKET.md shows the packet actually opens and
reads as text (# KRTV Studios — Design Packet…), not a corrupted binary.

Open WIREFRAMES.html. Expect: browser tab title KRTV Studios — Hi-Fi Wireframes and
page heading KRTV STUDIOS — WIREFRAMES.

Scroll the full page. Expect: exactly 10 frames, headed in order: 1a · Landing — doors
closed, 1b · 2 · Rates (doors open), 3 · Boarding Times, 4 · Checkout,
5 · Confirmation — MetroCard, 6 · Contact, 7 · Location, 8 · Sign in / Register,
9 · My Bookings, 10 · Admin — Schedule board — matching the 10-screen inventory in
PACKET.md §2.2.

Open devtools → Network tab, then reload. Expect: zero outbound requests — the file is
fully self-contained (verified by source inspection: no http://, https://, or external
src=/href= anywhere in the HTML).

Click the doorbay graphic on the Landing frame. Expect: it toggles open/closed — this is
the one scripted interaction in the file (doorbay click listener toggling doors-open).

On the Rates frame, click the WITHOUT ENGINEER tab. Expect: nothing happens — the tab
has no click handler. This is the correct behavior for a hi-fi mockup, not a bug: only the
doors are wired up; the WITH/WITHOUT toggle is presentation only until real code exists.

Final check: inspect the CSS variable --font-ticker — it resolves to "Courier New",
ui-monospace, monospace, with an inline comment /* stand-in for SubwayTicker.ttf */ — matching
NOTES.md's own caveat that the display font is a placeholder, not the production asset.

Read PACKET.md §2.1's digit-color mapping. Expect: 1→purple, 2/3→brown, 5→blue,
6→green, 0→yellow/orange, $→green, FOR→red.

On the Rates frame's three rows, read the badge classes in the HTML source (view-source:).
Expect: row 1 ($150 · 3 HOURS) is b-green b-purple b-blue b-yellow; row 2
($200 · 5 HOURS) is b-green b-brown b-yellow b-orange; row 3 ($350 · 10 HOURS) is
b-green b-brown b-blue b-orange — every digit's color matches §2.1's mapping exactly.

Check the FOR badge's CSS rule (.for{background:var(--line-red)...}). Expect:
var(--line-red) — confirming FOR→red is implemented, not just claimed.

Final check: the three WITH-ENGINEER prices rendered ($150, $200, $350) match
PACKET.md's own stated legacy prices for 3/5/10-hour sessions with an engineer, word for word.

Read PACKET.md's "Honesty caveat block". Expect: it states the without-engineer prices
as $60/3h, $100/5h, $350/10h and explicitly flags note the legacy $350 appears in BOTH
columns; likely a bug ⚠ — confirm the without-engineer 10-h price.

Read NOTES.md's open-question #3. Expect: the same ambiguity is restated as a named,
numbered blocker: Without-engineer 10-hour price — legacy shows $350 in BOTH columns,
almost certainly a copy-paste bug. What's the real price? (blocks rate seed, Gate 3).

Check WIREFRAMES.html's Rates frame for the WITHOUT-ENGINEER row values. Expect: they
are not rendered — only the WITH-ENGINEER tab is populated (its tab has class tab on;
WITHOUT ENGINEER has no row markup at all in the source) — so the ambiguous price was never
drawn into a screen where someone could mistake it for a resolved value.

Final check: attempt to independently verify the real without-engineer 10-hour price against
the legacy source. This cannot be verified — the legacy apps/client codebase is not on this
disk (per _handoff/krtv.md); the correct price remains genuinely unknown until Day 0 asks the
human.

Read PACKET.md §2.1's full token list. Expect: 12 named hex values — amber #fdd894;
flicker chain #ffec9e, #ffbb70, #ff992b; 8 MTA line colors #EE352E, #00933C,
#6CBE45, #0039A6, #FF6319, #FCCC0A, #B933AD, #996633; gray #A7A9AC; door gray
#808183; MetroCard gold #EFBB01.

Read krtv-prod-ready-plan.md Day 1 → AI3 and extract its literal G4 grep pattern. Expect
(quoted verbatim): -E '#fdd894|#EE352E|#0039A6|#FF6319|#FCCC0A|#B933AD|#996633|#6CBE45|#00933C'
— 9 of the 12 tokens named in step 1.

Run that exact pattern against the 3 tokens it omits:
printf '#A7A9AC #808183 #EFBB01\n' | grep -E '#fdd894|#EE352E|#0039A6|#FF6319|#FCCC0A|#B933AD|#996633|#6CBE45|#00933C'.
Per G4's own stated invariant in PACKET.md §1.6 (any occurrence of #fdd894, MTA line hexes,
or font-family: *Subway outside tokens.css/tailwind.config fails CI), the correct
behavior is that this command matches and exits 0, flagging the hardcoded duplicates.
Currently fails: it prints nothing and exits 1 — gray, door-gray, and MetroCard-gold
duplicates would silently pass CI. DEFECT — record as fail.

Final check: confirm this isn't a misreading — grep -n "A7A9AC\|808183\|EFBB01" PACKET.md
returns the same §2.1 token line for all three, proving they are genuinely declared tokens the
guard is supposed to protect, just missing from its own reference implementation.

Run grep -oE '\bG[0-9]+\b' PACKET.md | sort -u. Expect: G1 through G5 only, each
defined once in the §1.6 table as a domain/CI invariant (G1 no-double-booking,
G2 no-unpaid-confirmations, G3 price-integrity, G4 token-lock, G5 contract-drift).

Run grep -n "Gates index" -A2 krtv-prod-ready-plan.md. Expect: a second, independent
G0–G10 scheme, one per roadmap Day, where the label text for G4 reads
G4 payments guards — not token-lock.

Compare PACKET.md's G1 no-double-booking against the roadmap's Gates index label for
G1. Expect them to mean the same thing; they do not. The roadmap's G1 reads
G1 CI+token guard proven — i.e. the roadmap's "G1" is actually testing PACKET.md's G4.
Two documents in the same packet use the identical short label G1 (and G4) for two
unrelated concepts. DEFECT — record as fail; a gate-by-gate executor citing "G1" or "G4"
without naming which document could confirm the wrong invariant.

Final check: PACKET.md's own "Gated build steps" summary (bottom of the file) has exactly
8 numbered steps
(sed -n '/^# Gated build steps/,/^# Honesty caveat block/p' PACKET.md | grep -cE '^[0-9]+\.'
→ 8), while krtv-prod-ready-plan.md has 11 Day headers (Day 0–Day 10,
grep -c '^### Day' krtv-prod-ready-plan.md → 11). Step 8 of the packet's summary
("Hardening: reduced-motion, Lighthouse≥90, throttling, backups, deploy") silently folds
three of the roadmap's distinct gated days — Day 8 (testing/observability), Day 9
(deploy/ops, including the restore drill), Day 10 (docs/launch) — into one bullet with no
restore-drill or docs deliverable named. Confirming decisions off the packet's 8-step summary
alone is not the same as confirming the full 11-day roadmap.

Read plan-queue.md's re-planning kickoff prompt. Expect (quoted verbatim): Load
/mnt/user-data/outputs/krtv/PACKET.md, NOTES.md, WIREFRAMES.html, and
krtv-prod-ready-plan.md. Run ls /mnt/user-data/outputs/krtv/PACKET.md on this Mac.
Currently: ls: /mnt/user-data/outputs/krtv/PACKET.md: No such file or directory — the
prompt bakes in a Claude.ai sandbox output path that has never existed on this machine.
DEFECT — record as fail.

Read krtv-prod-ready-plan.md's own header line. Expect (quoted verbatim): companion
files named as krtv/PACKET.md, krtv/WIREFRAMES.html, krtv/NOTES.md — implying a nested
krtv/ subdirectory. Run ls ~/Projects/krtv/krtv/PACKET.md (the path its own header
literally implies, relative to a project root). Currently:
ls: /Users/wolf/Projects/krtv/krtv/PACKET.md: No such file or directory — the roadmap
assumes its own companions sit one directory deeper than where they will actually land once
unzipped. DEFECT — record as fail.

Final check: the one path that actually works is the flat one —
ls ~/Projects/krtv/PACKET.md (after unzip) succeeds and lists the file. Neither cross-reference
inside the packet states this correct path; a future executor must know to ignore both and use
the flat layout.

Count NOTES.md's "Named deferred list". Expect: exactly 10 numbered items (the
handoff and PACKET.md both describe it as a "10-item deferred list").

Count NOTES.md's "Open questions" list. Expect: exactly 12 numbered items (matching
every reference elsewhere in the packet to "the 12 open questions").

Search PACKET.md for any resolved-decision changelog (the roadmap's GATE 0 requires
user confirmations pasted verbatim into PACKET.md changelog). Expect: no such changelog
section exists yet — decisions D1–D6 in the "Decisions block" still carry their △/⚑/⚠
flags, none replaced with a resolved value.

Final check: cross-check against the Observify board — krtv has zero cards in in_progress,
testing, done, or deployed. The board, the disk, and the packet's own unresolved flags all
agree on the same fact: Day 0 (decision lock) has not been run.