LLL · communication board

Verification and Constraints

● READY FOR REVIEW

Kertújítók · 5,000 Ft/day launch and evidence-backed setup checks

Live owner: Parent · preparation Run: kertujitok-ads-5000-20260929 Updated: September 29, 2026 · 10:54 Budapest baseline

Goal and finish line

End goal: Run the six prepared Search campaigns at a combined average daily budget of 5,000 Ft and verify their setup.

Done means: The agreed budgets and enabled states have fresh provider readbacks. Every setup check has evidence or an explicit limitation. Configuration, eligibility, delivery and enquiries remain distinct.

0 / 10criteria verified

5deliverables

2open decisions

Rules that stay fixed

Preserve the 200,000 Ft whole-account calendar-month ceiling and unrelated paused campaigns.
Google Ads API only for account reads and changes. Cloudflare only. No GAJA/GHL, payment controls, client messages, paid self-clicks or duplicate campaigns.
No Ads changes during LLL Prepare. After this board is reviewed, a later explicit Launch starts one Sol Medium executor.

Proof standard: Fresh provider readbacks prove settings. GUI criteria require actual screenshots, independent descriptions and comparison. Missing evidence stays unchecked.

Budget and timing

Assumption: the same six prepared campaigns share 5,000 Ft/day in total. Allocation scales the existing priorities. It is not a performance claim.

Campaign	Current Ft/day	Proposed Ft/day
Soil drilling	249	1,386
Concrete breaking	235	1,308
Paving	180	1,002
Turf laying	62	345
Irrigation installation	62	345
Local garden maintenance	110	614
Total	898	5,000

Google uses an average daily budget. Most campaigns can be charged up to twice the highest daily budget set that day. Google budget rules.

Timing: the previous run is scheduled to pause September 29 at 17:25 Budapest. Prepare leaves that schedule in place. A reviewed Launch must reconcile the old pause before applying the new run. September 30 at 23:59:59 remains the native end unless Matt explicitly extends it.

Parent ↔ child messages

Newest message first

CHILD · Evidence baselineSeptember 29 · 10:54 Budapest

Confirmed by fresh Google Ads API reads: six enabled and eligible campaigns, seven approved active ads, other 19 campaigns paused. All six and the whole account still report zero impressions, clicks and cost for September through today. Three post-v49 form retests remain pending.

PARENT · Evidence baselineSeptember 29 · 10:54 Budapest

No executor has been created.

Parent read setSmall files first

After a stall or completion, read the clean gist plus these short files. Keep raw thinking and tool logs out of the normal review path.

Clean transcript gist Deliverable list Main prompts used

Work and proof

0 / 10 criteria verified

Confirm scope and budget

limit every mutation to the intended campaigns and affordable exposure.

0 / 2

Criteria, constraints and evidence

Observable tests

When the six IDs and all account campaign-budget associations are queried immediately before writing, then exact names, HUF currency, private/shared budget ownership, current statuses, budgets, dates and a restorable before-snapshot, and API-exposed account-serving restrictions or billing/payment-hold indicators appear in ROOT/evidence/preflight.json. When the proposed six budgets and whole-account month costs, daily high-water budget history and remaining dates are reconciled, then an exact5000HUF sum and a conservative exposure estimate below200000HUF appear in ROOT/evidence/budget-ledger.json, with reporting latency and overdelivery assumptions stated.

Budget proof must cover all account spend, delayed reporting, the highest budget set on the change day, and every remaining active date. State the numerical reserve and why it bounds uncertainty. Google's official rules allow most campaigns up to twice that day's highest daily budget, and an end-date edit counts as a budget change: https://support.google.com/google-ads/answer/10487143?hl=en. Do not treat 5,000 HUF as a hard daily charge ceiling. If a defensible upper bound cannot fit beneath 200,000 HUF, preserve affected existing values and report the exact missing evidence or shortfall. A read-only monitor cannot enforce a cap.

Protect

✕ Do not mutate a shared budget affecting unrelated campaigns.
✕ Preserve the monthly ceiling and exact campaign identity.
✕ Refresh if review delays make evidence stale.
✕ If the API cannot expose a billing indicator, record UNKNOWN without switching to the Ads browser or modifying payment controls.
✕ A known hold prevents a serving claim, not the separately authorized budget/status configuration.
✕ State the exact account action needed without taking payments into scope.

Pending execution. No criterion marked passed from preparation evidence.

Audit and repair campaign setup

establish what is correctly configured and repair narrow mistakes.

0 / 2

Criteria, constraints and evidence

Observable tests

When campaign settings, networks, location inclusion/exclusion and presence settings, languages, dates/schedules, bidding/CPC caps, groups, keywords/match types/negatives, ads/assets/policy/final URLs and conversion goals are inspected for the six IDs, then per-campaign PASS/FAIL/UNKNOWN findings, source readbacks and before/after evidence of authorized repairs appear in ROOT/evidence/setup-audit.json. When both maintenance groups199215721566 and202090510244 are inspected against their10km local service intent, then their individual targeting inheritance, ads, keywords and status evidence appear in that setup audit with both groups explicitly covered.

Recover the approved expected settings from the prior launch artifacts and current CURRENT_ARCHITECTURE.md before assigning PASS. Record expected value, source, observed value and comparison for each check, including geography, schedules, strategy, conversion action IDs, and service/ad/page alignment. Recover missing facts autonomously. Unrecoverable expectations remain UNKNOWN rather than becoming an invented strategy.

Protect

✕ Preserve business geography, approved service claims and existing strategy unless fixing an evidenced misconfiguration.
✕ No new services, duplicate campaigns, broad expansion, payment controls, or GAJA/GHL.
✕ Unknown API fields are not implicit passes.
✕ Matt explicitly requires: “use the google ads API, not the browser”.
✕ Perform every Google Ads read and mutation exclusively through the API.
✕ Do not open the Ads UI or require Ads UI screenshots.
✕ Browser evidence remains in scope only for landing pages and the review board.
✕ An API-inaccessible check remains explicitly UNKNOWN after supported API recovery.

Pending execution. No criterion marked passed from preparation evidence.

Verify landing and conversion paths

distinguish reachable pages from a working enquiry journey.

0 / 2

Criteria, constraints and evidence

Observable tests

When each active final URL is requested and its service/form path is tested with safe clearly marked synthetic inputs through the normal Cloudflare route, then URL/redirect/relevance results and the three outstanding post-v49 form retest results appear in ROOT/evidence/form-checks.json, individually PASS/FAIL/UNKNOWN with event/storage/duplicate checks where accessible. When desktop and mobile form paths are tested in the actual current interface, then dated screenshots, independent gpt-6-luna high descriptions without expected-result hints, and parent comparisons appear beside the corresponding criterion on the board. Missing screenshots remain unverified and do not alone block otherwise provable launch.

Before testing, recover the three post-v49 retest scenarios from existing project evidence and create a three-row mapping of source, route, scenario, expected event/storage/duplicate outcome and result. Do not substitute three generic submissions. If a scenario cannot be recovered, identify the missing specification explicitly.

Protect

✕ Never bypass existing CAPTCHA/consent, create fake ad clicks or claim a platform event is a real lead.
✕ Suppress outbound notifications for synthetic QA using an existing supported path and clean only identifiable task test records.
✕ If normal testing would message a client, block only that test and record it.
✕ No new security layer or unrelated site redesign.
✕ Pending form proof alone does not revoke Matt's explicit launch authority.

Pending execution. No criterion marked passed from preparation evidence.

Screenshot proof pending: actual tested interface → independent Luna high description → parent comparison. Keep unverified until all three exist.

Apply budget and enable campaigns

make the requested spend settings real and recoverable.

0 / 2

Criteria, constraints and evidence

Observable tests

When the reviewed allocation is applied only to the six owned budget resources and their campaign statuses, then a separate fresh Google Ads readback shows budgets1386/1308/1002/345/345/614HUF and ENABLED status for the mapped IDs in ROOT/evidence/launch-receipt.json, with every unrelated campaign ID/status matching the immediate before-snapshot and agreed dates unchanged. The earlier count of 19 paused campaigns is context only. Report concurrent unrelated changes without overwriting them. When the old48h pause automation is reconciled before activation and the replacement observation is saved, then effective automation state and next observation time appear in ROOT/evidence/automation-receipt.json without an obsolete scheduled pause capable of undoing this launch.

Protect

✕ No mutation until later explicit Launch under LLL.
✕ Keep narrow rollback snapshots and never claim enabled means serving.
✕ If provider rejects a partial mutation, reconcile each entity and restore affected values or finish within the reviewed total before claiming success.

Pending execution. No criterion marked passed from preparation evidence.

Publish verified findings

make outcomes and remaining issues reviewable.

0 / 2

Criteria, constraints and evidence

Observable tests

When impressions, clicks, cost, policy/account restrictions and actual conversion evidence are queried after the changes, then timestamped results and a grounded diagnosis of any zero delivery appear on the board, separating verified causes from hypotheses. When the updated board is fetched with an unauthenticated client and opened in the in-app browser, then the stable bare review URL returns200 without login/token redirects and shows current checked criteria, evidence summaries, constraints and unresolved items. ROOT/evidence/publication.json records HTTP proof, with actual render screenshot and independent description required for a GUI-render pass.

Protect

✕ No client messages or automatic audio publication.
✕ Keep credentials, private raw account evidence and personal data off the public board.
✕ Preserve stable document identity/comments and existing unsent email state.
✕ Do not green-check missing evidence or invent delivery.

Pending execution. No criterion marked passed from preparation evidence.

Screenshot proof pending: actual tested interface → independent Luna high description → parent comparison. Keep unverified until all three exist.

Open decisions and blockers

Only unresolved items

Scope and end-date assumptionsAwaiting clarification

Two questions were sent: confirm these six at 5,000 Ft/day combined, and retain September 30 or continue beyond it. Pending an answer, this draft uses six campaigns, 5,000 total, September 30 end. Zero delivery remains unexplained. A budget increase is not proof of a fix.

Unverified items remain visible above.

Recovery state

Only inspect on stall or completion

0consecutive stalls

Not startedrecovery state

Review boardresume point

Last stall / completion note: Preparation only. No executor, budget mutation or relaunch has occurred.

Parent action: Review this board, resolve any changed scope/date assumptions, then send Launch. The executor must check latest comments and inline rewrites before launch.

Execution authority and completion

At execution, save the later Launch instruction and the exact displayed contract revision or SHA256 it covers. It covers the visible six-campaign combined-budget and September 30 default assumptions unless Matt corrects them. Do not request the same authorization twice or create a new confirmation gate. Read any newer user correction and board comments before writing. After LLL's later Launch: refresh provider evidence, apply the agreed5000HUF split and enable these six, repair narrow evidenced setup defects within the above constraints, reconcile the obsolete pause, create a bounded observation, maintain this board. Routine reversible recovery is authorized. Payments, messages, broader campaign scope, date extensions without explicit direction and unrelated production work remain outside scope. No independent executor is created during Prepare. The future named child is one pinned Sol Medium executor and owns implementation and any permitted internal workers. Keep deliverables.md, main-prompts.md and session-id.txt as the short parent read set.

Complete only when required outcomes are proven and the board agrees. A completed inspection with FAIL or UNKNOWN results is not proof that all campaigns are set up properly. Distinguish budget/status execution complete, audit complete with limitations, and full requested outcome complete. Unknowns remain unchecked. If an essential capability is irreducibly unavailable, return the exact affected blocker and resume point, not 'all set up properly'. An unresolved delivery diagnosis may coexist with verified configuration but must be disclosed.

Reviewed contract SHA256: 056a7bffb27e47814bcf5709b369162e603f9ef2550d8e4c7c7d6fefab6ed484
Review coverage: Astra low synthesis plus Sol low and Luna max reviews. Required Luna xhigh is unavailable. Independent external-assumption review is incomplete. Official budget rules were checked directly.