feat(linkedin-studio): N15 — do-next-kontrakt (recordDoNext + skrivere/lesere + forrige-utgave-kalibrering) [skip-docs]

Måling som ikke endrer neste utgave er teater: report/analyze/ab-test/48h-monitor
endte alle i chat (grep do.next = 0 i hele repoet). N15 gjør kanalen til en kontrakt.

- recordDoNext i hooks/scripts/state-updater.mjs (TDD, 17 nye tester): seksjonen
  Do-Next Directives, rader "[dato] (kilde) direktiv — evidence: peker".
  Levetid = erstatt-per-kilde + 60-dagers aldersgulv, maks 3 per skriv, additiv
  scalar last_donext_date, $-sikre replacement-funksjoner. CLI: --record-do-next.
- Skrivere: report Step 7b, analyze Step 6b, ab-test 2c.6 (kun Adopt),
  post-feedback-monitor (kun 48h-sjekken); analytics-interpreter emitterer nå
  anbefalinger på do-next-form (imperativ + evidens).
- Lesere: post/quick/batch/create Step 0 + newsletter Step 1, som i tillegg
  kalibrerer på forrige utgaves MÅLTE tall via queue-id-match (queue.json →
  analytics/posts) med eksplisitt honest-miss — aldri estimert.
- Guards: test-runner Section 16v (15 ubetingede sjekker, floor 198 → 213),
  $-safety-batteriet dekker recordDoNext (coverage-gate fanget den nye eksporten).

Suiter: test-runner 232/0, hooks 191/0, trends 300/0, brain 134/0, editions 72/0,
specifics-bank 45/0, contract-gate 33/0, tests 35/0, render 60/0.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QxvWAjte7vPcF79QeSRvRJ

agents/analytics-interpreter.md

@@ -343,9 +343,9 @@ Based on this analysis, your unique advantages are:
 - Followers: [↑/↓/→] [net change]
 
 ## Recommendations for Next Week
-1. [Most impactful action]
-2. [Second priority]
-3. [Experiment to try]
+1. [Most impactful action] — evidence: [the measured number it rests on]
+2. [Second priority] — evidence: [measured number]
+3. [Experiment to try] — evidence: [measured number]
 
 ## Content Plan Adjustment
 Based on this week's data:
@@ -354,6 +354,14 @@ Based on this week's data:
 - Start: [new experiment]
 ```
 
+**Write recommendations in do-next form — imperative + evidence pointer.** The calling command
+persists the top 2–3 verbatim as do-next directives (`recordDoNext`, source `report`), and the
+create surfaces read them at Step 0 before drafting. So each one must be actionable by someone
+who never saw this report: "Plan 2 AI-agent posts this week" (imperative, checkable), not
+"AI content is performing well" (an observation nobody can execute). The evidence pointer is
+what lets the next session judge how much weight a directive carries — always name the measured
+number and the sample it came from.
+
 ### Monthly Report Additions
 
 For monthly reports, also include:

agents/post-feedback-monitor.md

@@ -287,6 +287,33 @@ Output a structured intervention plan using this format:
 - [Actionable change to try next time]
 ```
 
+### Persist the 48h learning as a do-next directive
+
+At the **final check (48h / Resurrection Window)** — not at the intermediate check-ins, where
+the numbers are still moving — write the one or two learnings that should change the next post
+into the shared do-next channel. This is the same channel `/linkedin:report`,
+`/linkedin:analyze` and `/linkedin:ab-test` write to, and the one every create surface reads at
+its Step 0. Without it the 48h learning lives only in this session's chat and is re-learned
+from scratch next week:
+
+```bash
+node --input-type=module -e "
+import { writeState, recordDoNext } from '${CLAUDE_PLUGIN_ROOT}/hooks/scripts/state-updater.mjs';
+writeState(content => recordDoNext(content, {
+  recordDate: 'YYYY-MM-DD',
+  source: '48h-monitor',
+  directives: [
+    { directive: 'Put the concrete number in the first line, not the third', evidence: '48h on 2026-05-28 post: velocity 82/100, golden hour 3x the account average' }
+  ]
+}));
+"
+```
+
+Write only what a *future draft* can act on — a per-post tactic ("reply to every comment inside
+the first hour") belongs in the report to the operator, not in the channel that shapes the next
+draft. `recordDoNext` replaces this source's previous row on each write, so the most recent 48h
+learning is the one in force.
+
 ## Step 7: Follow-Up Scheduling
 
 Based on current performance, suggest: