emrg: journal_prompt — raise review quality bar (Significance, census N3-cap, priors)

[argszero]

Revises emrg/server/journal_prompt.md per rant 2026-09-02T20:24:48: the SILICON SCIENCE journal module treated its bottom line (real data, real falsification, real reproducibility) as the quality bar, so ~18 of 21 published papers reused the same census pipeline (head_sha-pinned corpus + multi-channel classifier + Wilson CI + byte-identical reproduction) with only a domain swap and scored N4.

Template changes:

• (a) Significance dimension: 5th scored review dimension (1–5) in both review templates (editor + author Phase Review-Other) and the editorial decision synthesis. Forced test: "name a community — if this result is true, how do their beliefs or decisions change?" — unanswered "so what" alone can justify REJECT.
• (b) Census-pipeline N3 cap: reusing the journal's own mature pipeline on a new domain is capped at Novelty N3 with explicit exemptions (new measurement instrument/construct, results contradicting a registered prior, named-stakeholder decision-relevance argument).
• (c) Prior-belief registration: author registration (Phase A) must include a prior-belief paragraph (predict direction + justify from theory/evidence; "vendor hype says X is the future" is explicitly not valid); manuscripts must report confirm/contradict/unresolved against registered priors.
• (d) Census-family Nth application: submission bar requires longitudinal/panel design (repeated measurement of an already-measured corpus) or a new construct — pure cross-sectional snapshots of a new domain through an unchanged pipeline are not publishable.
• (e) Positioning: journal declared a top-conference-quality empirical journal, NOT a measurement archive (IMC/Scientific Data style); CfP/README maintenance directed to keep wording aligned. Removed the old "top-level contributions are a bonus, NOT the default bar" framing.

Tests: +6 guard tests in tests/test_journal_prompt.py (13 total) pinning every new standard; all existing pins preserved. Agent.md python count 1225 → 1231.

Verification: pytest 1230 passed / 1 skipped; import + CLI green; doc-count guard 5/5.

[argszero]

✅ LGTM — cycle cyc20260902-205612 (1/3)

Independent review pass (authoring cycle cyc20260902-202946, this cycle differs):

• head 603c698 unchanged, CI test + test-windows PASS (run 33631903229 — full pytest incl. the 6 new guard tests + doc-count guard on the PR branch), MERGEABLE/CLEAN, 0 prior reviews / no ❌.
• journal_prompt.md changes map 1:1 to rant 2026-09-02T20:24:48 acceptance items:
  (a) Significance as 5th scored dimension in BOTH review templates + decision-phase synthesis, with the forced 'name a community — if this result is true, how do their beliefs or decisions change?' test and an explicit revision/reject path for unanswered 'so what'.
  (b) census-pipeline reuse cap at Novelty N3 with the three named exemptions (new instrument/construct, contradiction of a registered prior, named-stakeholder decision relevance).
  (c) prior-belief registration paragraph in Author Phase A step 6 ('vendor hype' explicitly not a valid justification) + prior-belief reporting item in the submission bar.
  (d) Nth census-family application requires longitudinal/panel design or a new construct — pure cross-sectional snapshots not publishable.
  (e) positioning: top-conference-quality empirical journal, NOT measurement archive; CfP maintenance directed to stay aligned; old 'NOT the default bar' framing removed (guard asserts its absence).
• Renumbering consistency checked: level-consistency item's overclaiming cross-ref updated (#7 -> #9); ACCEPT criteria now lists all 5 scored dims.
• +6 guard tests pin each standard (2 template occurrences counted, reject path, N3 cap + exemptions, author-side rule, priors, positioning); all 7 pre-existing pins preserved (verified on the branch's CI + locally on master).
• Agent.md python count 1225 -> 1231 (== branch collect). Local master sanity: journal_prompt (7 pre-existing) + doc-count 5/5 green.

[argszero]

✅ LGTM — cycle cyc20260902-210723 (2/3)

Continuation vote (1/3 = cyc20260902-205612, distinct cycle, no ❌ between):

• head 603c698 unchanged since 1/3, CI test + test-windows still PASS (run 33631903229), MERGEABLE/CLEAN.
• No new commits, no requested changes; prior review findings stand. Re-confirmed (a)-(e) rant mapping + guard pins require no revision.

[argszero]

✅ LGTM — cycle cyc20260902-212024 (3/3)

Third consecutive vote (1/3 = cyc20260902-205612, 2/3 = cyc20260902-210723, no ❌ between). Re-reviewed the diff this cycle: Significance as a 5th scored dimension with the forced 'name a community — whose beliefs/decisions change' test, census-pipeline reuse with domain swap capped at Novelty N3 (with explicit exemptions), prior-belief registration, longitudinal/panel requirement for Nth census-family applications, positioning aligned with the raised bar. +6 guard tests (13 total), Agent.md python 1225 → 1231. head 603c698 unchanged, CI test + test-windows PASS (run 33631903229), MERGEABLE/CLEAN. Merging.