@stage5/lumine 0.3.68 → 0.3.70

package/sdk/LUMINE_ADMIN.md

@@ -1186,6 +1186,7 @@ lumine admin report serve --file <dir>/daily-management-report.md [--port <n>] [
 ```
 
 These compose the documented commands in this contract, running them as child processes, so every scope check, audit and pagination rule applies unchanged.
+- **Arena members waiting for papers come first.** `arena paper-queue --cursor <member-id> --limit <1-100> --json` lists members blocked by the lack of a fresh complete published paper, using the member-facing pass/retry and solo-mastery rules. Follow every `nextCursor`, including after an empty page; require `demandBasis: waiting-published-papers-v1`. Group all pages by exam/difficulty and prioritize these waiting members before generic unseen-question depth. Failed ordinary papers remain retryable; open/pending-marking sittings are not publication waits. Ten successful distinct Full Hard clears, with Zero/Ciel allowed, are needed between solo mastery attempts. The pass and full-marks 20× XP/Coins rewards are separately reserved, each once. Publish complete independently reviewed papers and verify the same members can advance. `gather` collects this queue first as `arena-paper-queue-N.json`, with `paperQueue.complete/pages/waitingTracks`; missing/malformed/partial coverage must not be reported as an empty healthy queue. Use the workspace CLI until npm publication is verified. This step grants no new authoring spend or publication authority.
 - **`gather`** saves its fixed set of read-only reports into `<dir>`, including the canonical published Math Lab queue in `math-lab-queue.json` and all types of pending Build review requests in `build-reviews-N.json`. Build reviews follow `nextCursor` to exhaustion; `gather.json` → `buildReviews` records `complete`, attempted `pages` and collected `items`. A failed/malformed page, cursor loop or 100-page bound reports incomplete coverage, never an empty healthy queue. Read the current collection once; it replaces the rewards-only list. Older packages without `buildReviews` need a separate `review list --status pending` collection through cursor exhaustion. Previous queue receipts remain on disk as history but do not become current evidence or duplicate reading duties after a new gather. It is not the complete daily checklist: app refills/usage, individual review assessments, sponsor integrity and other duties need their own steps. Repository agents follow the current full-run coverage index in `agent-guides/lumine-admin.md`. It follows `bot-output` to exhaustion and renders `bot-chats.txt`, then lists any failed step in `gather.json`. Total/per-step `durationMs`, attempts and recovered reads make recurring slow or failing collection steps visible during the daily review. It also creates `review-state.json`: collection never acknowledges reading. Children bypass update checks, require matching fresh stdout/file receipts, and retry an unconfirmed opted-in read at most once. Mutations never auto-retry. It needs the active run for run-scoped reads.
 - **`progress`** reads local review state and explicit operator acknowledgments, writes private `progress.json`, and keeps reading, canonical effort mutations, the run lease, and browser checks separate. `collectionFailures` includes failures without a page file, such as a cursor loop; acknowledging the available files cannot make an incomplete collection complete. A fresh successful gather clears those failures. Add all required duties for the actual run scope and its carryovers; collection alone is not a full review. A completed run lease cannot complete a pending duty. Reading acknowledgments bind to the current file bytes and are operator attestations, not proof of comprehension.
 - **`featured candidates`** returns Subjects posted after the Bangkok-midnight cutoff that are off the board and provably never Featured. Each one comes with its human/bot comment counts, explicit comment truncation/completeness, and whether Mikey has viewed it. Counts from truncated context are partial. An unavailable Subject context fails the collection. Eligibility only: ranking stays editorial.
@@ -1880,6 +1881,10 @@ and seeding coverage for steps 1-3 once the app is approved.
   decisions. The old Math Studio steps 1–3 / 20-per-run recipe is historical,
   not current authorization.
 
+### Game diagnostic delivery
+
+For retained game delivery evidence, `lumine admin game-diagnostics summarize --file <envelope.json> --json` is local-only and requires no login. The envelope is `{buildId,topic,entries,coverage:{complete,since,until}}`; keep the canonical sharedDb `id`, `userId`, `topicId` and parsed `data` on every row. The command deduplicates stable event/batch IDs within their canonical user/topic/artifact scope, flags conflicting copies, and reports legacy/unknown-version coverage. Raw evidence is still required for judgment. Cumulative delivery snapshots are not summed; server collection completeness does not prove complete client delivery. The reviewed October28 producers and CLI are not live until their release receipts confirm them. See `work/games-followup-20261028/` and the repository's daily guide.
+
 ### Reward activity report (standing duty, every full daily review; added 2026-09-12)
 
 Also inspect failed or unconfirmed payouts with `lumine admin reward-diagnostics --days 1 --json`.
@@ -3155,6 +3160,46 @@ type PostSkip = Success<{
 
 ## Twinkle Newspaper
 
+**Daily website-code update article (Mikey, 2026-10-28):** before claiming an
+edition, the duty bot (Zero/Ciel) publishes a real Subject at reward/effort level
+**2** about member-relevant website code changes shipped since the previous
+paper. Check Git/push history against successful production deployments and the
+actual live revision, then explain the useful changes in plain language; a push
+or a proposed fix is not a release. This is website development news, not a
+roundup of member posts or internal tooling. Cite that Subject's actual event
+from the claim digest and include the linked article on every front page; use
+it as the lead only when its significance warrants it. Verify author, level,
+source link and printed placement, and avoid duplicate posts or repeated release
+coverage. If no relevant changes shipped, say so briefly without inventing news.
+The standing instruction authorizes this scoped post and coverage; it grants no
+unrelated release authority. Keep release evidence and private diagnostics out
+of the public article. Follow the repository's daily website-code update
+procedure and the existing dated-revision contract for an approved correction.
+
+Use the scoped posting command after composing and reviewing the article:
+
+```bash
+lumine admin subject post-update --file article.json --json
+lumine admin subject post-update --file article.json --confirm --json
+```
+
+The first command is a read-only preview. The second publishes as the selected
+run's duty bot at level 2; `newspaper` or `full` run scope is required. The file
+contains `title`, `description`, `coverageStart` and `coverageEnd` (Unix seconds),
+and `releaseEvidence`: an array of `{repository, commit, deployedAt, summary}`.
+Use full Git commit hashes and verified deployment times inside the completed
+coverage interval. The evidence stays private; put only useful member-facing
+information in the article. An empty evidence array is allowed when no relevant
+changes shipped, and the article should say that briefly. The command does not
+independently verify GitHub/Vercel receipts: the operator must do that first.
+
+One canonical article is permitted per scheduled editorial day. An identical
+retry returns its existing Subject instead of notifying members again; a
+changed payload, changed duty bot or edited/deleted post requires inspection
+and is not silently overwritten or recreated. Preview reports any existing
+Subject. The already-published October28 article is reserved as a legacy post;
+reuse Subject37265 rather than trying to republish it with this new command.
+
 **Audience (Mikey, 2026-09-15): children and young Twinkle users aged 10–15.**
 Choose and write stories for what these readers would voluntarily spend time
 reading. Before selecting a story, identify its appeal to them: curiosity,
@@ -5467,6 +5512,10 @@ Legacy aliases such as `subjects list`, `subjects get`, `subjects featured`,
 `comments get`, and `recommend` remain accepted, but the singular command forms
 shown above are the canonical interface.
 
+For bounded case evidence, use `lumine admin jev-routing review --since <UTC-day> --until <next-UTC-day> --json` and follow every cursor, or `--all --checkpoint <persistent-file>`. Daily gather retains every20-event source page for the preceding complete UTC day. The reader covers paired disagreements and Auto fallbacks once per event; synthetic traffic is excluded unless `--include-synthetic` is explicit. It returns only currently unexpired frozen input and labels absent/expired input. A missing endpoint is unavailable coverage, not an empty healthy day. Collection never acknowledges reading or proves classification accuracy. This October28 reader and independent Auto-input capture are local until the matching release is verified.
+
+Before calling a reply unrelated, check its original generation trigger and the bounded `askedAbout` descriptor. A member may have edited the visible turn or asked about linked content. The descriptor explicitly represents current stored metadata, not a separately frozen generation snapshot; never bypass deleted context.
+
 Read `jevRoutingShadow` alongside `jevPilot` in both the admin brief and full
 daily report. The September 21 primary-chat experiment is historical: chat
 routing is off by default and served-route audits were retired September 23.