The user authorized the Campaign 1 run sheet for the September 8 capital window. Friday September 4 is the settled corpus; Monday September 7 is the Labor Day holiday. The clean analyst's result is one BUY (MU), fifteen WAITs and twelve explicit capital-window AVOIDs. This is a research plan with levels, not sizes or trades.

Phase checkpoints

Entry-map table, verbatim

Commission ledger

All commissions carry campaign-01: capital deployment window 2026-09-08. State transitions were made through the canonical emitter; none was simply hand-edited to completed status.

Twenty-one perspective questions remain needs-evidence; four phase umbrellas and the one MU company dive are emitted. The new company memo resolves MU-specific contract proof, not peer-wide inventory/contract durability. No additional company dive was needed for the admitted slate; the cap was eight, the actual count one.

Trust command and interpretation

python3 research/issues/2026-09-06-campaign-01-trust-receipts/call-metrics.py --input research/issues/2026-09-06-campaign-01-trust-receipts/desk-calls.json

30 days: 3 of11 eligible published BUY calls beat SPY (27.272727%, displayed27.27%); eight misses average10.37 percentage points behind SPY; nine BUY calls excluded.90days:zero eligible, 20 excluded; hit rate and mean miss unavailable. These are publication-date raw-price outcomes, not returns on fills or a test of entry-trigger execution. Supersession preserves the historical published stance. Campaign 1 itself is ungraded.

What was not done, and why

• No trades, account sizing, live-order work, push, deployment, schedule or new engine guard. The user authorized research/publication, with explicit no-trades and blind-input boundaries.
• No direct ledger/holdings reads. Daily money-loop/persona/autotrade steps that require them were excluded, so the campaign does not falsely report every due daily row cleared. Paper performance uses derived aggregates; the August 12 ladder is stale and cannot establish a fresh promotion. Conditional alerts and mechanical refreshes are not agent verdicts.
• No repeat Friday full scan: its final audit was already green with waivers. Provider/proxy/history holds, inaccessible premium financial ratios, source-link redirection and thin/ambiguous social results remain visible. Fresh retrieval dates do not change issuer publication dates.
• No future-dated Tuesday flagship: the canonical conditions snapshot rejects a future date. The published Sunday brief carries the map; the Tuesday Read is prepared as an internal handoff. The actual Tuesday edition still requires its dated source check and required flagship chapters. Nothing is scheduled.
• No historical hit-rate claim for the new campaign and no 90-day metric without eligible outcomes. No point-in-time counterfactual claim that September's process would have prevented every July loss.

What the run sheet got wrong, omitted or placed out of order

The campaign Anatomy is amended before Campaign 2. The principal corrections are: establish a reviewed source boundary before intake; build the complete focus/cohort manifest before spending news quota; separate current-RSI nomination from historical flag admission; inventory published calls as well as targets before Names; preserve original stances and earlier recorded outcomes when consolidating; distinguish entry hold, pattern failure and long-trend invalidation; reconcile same-period financials and cash-flow composition; define scoreboard denominators and missing windows; resolve no-ledger/no-trade conflicts explicitly; prepare future flagship prose without bypassing the actual-date guard.

Three contexts encountered legacy account annotations, including the coordinator. Their contaminated market analysis was stopped or excluded; the valid Phase 0/1 findings preceded the later analyst exposure, and all Phase 2/3/entry-map judgments came from a fresh read-only analyst using individually reviewed capsules. This is a real procedure failure, not a claim that merely omitting ownership from the final prose repairs contamination. The existing engine task list records the source-boundary, board vocabulary, DHR conjunction, call lifecycle and generated MU report defects. No mechanism fix or new gate is bundled.

The first brief validation also required a watch-shaped section; the final Sunday Read includes What to Watch. The 21 evidence gaps remain open decisions for later research, not an excuse to sweep additional work now. Two manual campaigns still precede any decision about automation.

Final publication and verification receipts

Campaign 1 is published. Command and output receipts record the actual checks, including the rendered table and board review.

The price checker uses an inclusive comparison: SNDK at 1740 produces an in-zone flag, while the authored WAIT requires strictly above 1740 plus a new flag. A price flag verifies only one leg and does not promote it to BUY. The existing engine task records this conjunction class alongside the DHR witness. The board-read vocabulary mismatch also remains filed: the research producer writes confirm, while the current UI accepts endorse; positive-review annotations may therefore be omitted by that consumer. The public entry map is the complete decision artifact. The /calls hub is prerendered at deployment and may retain its previous build; publishing the data mirror does not deploy code.

Final runlog triage: the two failed lint invocations are known, each tied to this report in ops/runlog-known.json. One strict source-lint run caught the report's code-case wording and a plain helper receipt being treated as an article; those campaign issues were corrected, and 328 pre-existing advisory bindings remain. The other caught two planned desk-thread references being interpreted as existing files; both now explicitly say planned, and strict path lint and closeout passed. runlog --failures then exited0 with nothing new. health --quiet remains nonzero because it retains the earlier strict-lint failure plus provider/history/heartbeat advisories; this is not an all-green health claim.

The Friday sample found ASML/MRVL/ARM percentages explicitly labelled above the 200-day being classified as daily returns, RTX's labelled weekly return classified as daily, and the English word “A” treated as a ticker in a crude paragraph. Those are sampled binding errors, not proof that all 328 advisory warnings are false. No acceptance baseline changed. Cadence still reports five overdue loops; money loops remain excluded under the campaign boundary, the curated sweep was not fired, and the nightly heartbeat retains its prior degraded status. These omissions are recorded rather than silently marked current.

The canonical publication command emits one unrelated account-metadata warning during its mechanical projections. The research receipt replaces that line with an explicit omission marker. This does not erase the earlier input-boundary failures described above or claim that the unfiltered command log contains no account reference. No such output was passed to the clean analyst or used for market judgments.

Sources

• Campaign authorization and amended Anatomy.
• Phase artifacts and code-linked receipts above; original primary URLs appear in the Climate, Theses, Names, MU and entry-map source footers.
• Market price authority: research/market-engine/data/summaries/*.json, September 4 settled; canonical daily-reader extracts archived beside the Names review.