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

commands/ab-test.md

@@ -342,6 +342,32 @@ handful of posts per variant a creator realistically gathers (well under the
 
 Update the test file status from ACTIVE to COMPLETED. Add the conclusion and recommended actions to the file.
 
+### 2c.6: Persist an Adopt verdict as a do-next directive
+
+An **Adopt** verdict (and only Adopt — "Keep A", "Inconclusive" and "Extend test" change nothing
+about the next draft) is a decision the next post must inherit. Write it to the same do-next
+channel `/linkedin:report` and `/linkedin:analyze` use, so every create surface picks it up at
+its Step 0 instead of the test result sitting unread in a file:
+
+```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: 'ab-test',
+  directives: [
+    { directive: 'Open with a question hook by default', evidence: 'test hook-form (6 posts): B +38% comments, directional-strong' }
+  ]
+}));
+"
+```
+
+Carry the **directional** framing into the evidence pointer (see 2c.4 — organic volume rarely
+reaches significance); a directive that reads as proven when the signal was weak is how one
+outlier becomes a permanent rule. `recordDoNext` replaces this source's previous adopt row, so
+a newer verdict supersedes an older one automatically — and the `report`/`analyze` slots are
+untouched.
+
 ---
 
 ## Step 2d: Review Test History

commands/analyze.md

@@ -220,6 +220,31 @@ Quick triage if a full audit can wait:
 - [ ] Monitor metrics closely
 - [ ] Continue high engagement activity
 
+## Step 6b: Persist the plan as do-next directives (the measurement→creation contract)
+
+The recovery plan above is worth nothing if the next drafting session never sees it. Turn the
+**2–3 load-bearing corrections** into do-next directives — the same channel `/linkedin:report`,
+`/linkedin:ab-test` and the 48h monitor write to, and the one every create surface reads at its
+Step 0:
+
+```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: 'analyze',
+  directives: [
+    { directive: 'Keep every post inside the 3 core pillars for 14 days', evidence: 'diagnosis: profile-content mismatch, reach -55% since off-topic run' },
+    { directive: 'No external links in the post body until reach recovers', evidence: 'diagnosis: link-in-body correlates with the drop window' }
+  ]
+}));
+"
+```
+
+Write the corrections that change the NEXT draft — not the whole checklist. `recordDoNext`
+replaces this source's previous rows and expires anything older than 60 days, so a superseded
+diagnosis cannot keep steering drafts. Confirm in one line what was persisted.
+
 ## Step 7: Timeline Expectations
 
 Set realistic expectations:

commands/batch.md

@@ -24,6 +24,10 @@ You are a LinkedIn batch content creator. Help the user create an entire week's
 
 Load state and personalization:
 - Read `~/.claude/linkedin-studio.local.md` for recent topics and weekly goals
+- **Read `## Do-Next Directives` in the same file** — the last measurement pass's directives
+  (report / analyze / A/B Adopt / 48h monitor) with their evidence pointers. A batch is where
+  they bite hardest: apply them across the week's posts and name which post carries which
+  directive in the plan. Never edit or clear the section (lifetime is managed per source)
 - Read `${CLAUDE_PLUGIN_ROOT}/skills/linkedin-studio/SKILL.md` for profile and preferences
 - Check `${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/plans/` for existing weekly plan
 - Read `${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/templates/my-post-templates.md` for proven templates — vary templates across the batch for format diversity

commands/create.md

@@ -26,6 +26,10 @@ If `~/.claude/linkedin-studio.local.md` exists, you may glance at the planned ne
 topic and recent pillars to make a smarter suggestion — but keep it to one line and
 do not block on it.
 
+If `## Do-Next Directives` in that file is non-empty, mention the top directive in that same
+line ("last report says: lead with the number") — it often decides which format to route to.
+You do not apply it here; the command you route to reads the section itself at its Step 0.
+
 ## Step 1: Identify what they want to create
 
 If the user's message already names a format (e.g. "a carousel about X", "react to

commands/newsletter.md

@@ -371,6 +371,39 @@ what genuinely changes the work.
 >    discovery→production loop automatically at Step 10 (scheduling flips the trend to
 >    `acted`).
 
+> **Calibration from measurement (A2-F14 — the do-next contract).** Two reads, both
+> cheap, both BEFORE the dimensions below. Neither asks the operator anything.
+>
+> 1. **`## Do-Next Directives` in `~/.claude/linkedin-studio.local.md`.** What the last
+>    measurement pass (`/linkedin:report`, `/linkedin:analyze`, an A/B Adopt verdict, the
+>    48h monitor) decided the next piece should do differently, each row carrying its
+>    evidence pointer. Apply the ones that apply to long-form and **name them in the brief**
+>    — a directive that shaped this edition must be visible in the brief, not silently
+>    absorbed. Never edit or clear the section here (each source manages its own rows).
+> 2. **The previous edition's MEASURED numbers — queue-id match.** The prior edition of
+>    this series was registered at its Step 10 under the stable queue id
+>    `<series-slug>-<NN−1>`. Resolve its numbers:
+>    - read `${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/drafts/queue.json`,
+>      take the entry whose `id` is `<series-slug>-<NN−1>` → its `scheduled_date`,
+>      `hook_preview` and `status`;
+>    - read the analytics batches in
+>      `${LINKEDIN_STUDIO_DATA:-$HOME/.claude/linkedin-studio}/analytics/posts/*.json` and
+>      select the `.posts[]` row whose `publishedDate` equals that `scheduled_date`
+>      (a rescheduled edition may be a day or two off — widen to ±2 days and disambiguate
+>      on a `title` that matches `hook_preview`);
+>    - carry forward `impressions`, `engagementRate`, `comments` and `saves` (when present).
+>
+>    Read them as calibration, not as a verdict: **low impressions** points at hook /
+>    topic-relevance (a Step 9 problem for this edition), **decent impressions with weak
+>    comments** points at resonance (a Step 6 persona problem), and a strong prior edition
+>    says keep the structure that worked. State the comparison in one line in the brief.
+>
+> **Honest miss.** No queue entry, no analytics match, or the edition simply is not measured
+> yet (`status: scheduled`, or the import has not been run) → say so in one line — *"forrige
+> utgave er ikke målt ennå; kalibrerer på skjelett og persona alene"* — and continue.
+> **Never estimate** the previous edition's numbers, and never let a missing measurement
+> block the brief. This is edition NN=1 of a new series' normal state.
+
 Settle these dimensions (most should come from context, not questions):
 
 - **Angle** — the one premise this edition argues.

commands/post.md

@@ -26,6 +26,14 @@ First, load persistent state and personalization:
 - Read `~/.claude/linkedin-studio.local.md` for posting state (streak, weekly progress, recent topics)
 - Read `skills/linkedin-studio/SKILL.md` for user profile, voice settings, and preferences
 
+**Read `## Do-Next Directives` in the state file and apply them to this draft.** This is what the
+last measurement pass (`/linkedin:report`, `/linkedin:analyze`, an A/B Adopt verdict, the 48h
+monitor) decided the next post should do differently — each row carries its evidence pointer.
+Apply the ones that fit this post, and **name the applied directive in one line** when you
+present the draft ("applied: lead with the number — report 2026-W22"). A directive that does not
+fit this topic is skipped silently, not argued with. Directives are already lifetime-managed
+(each source replaces its own; >60 days expire) — never edit or clear the section here.
+
 Check state for topic planning:
 - Compare intended topic against "Recent Posts" in state file
 - If a similar topic was posted in the last 7 days, suggest a different angle or topic

commands/quick.md

@@ -27,6 +27,10 @@ Read `~/.claude/linkedin-studio.local.md` for:
 - Weekly posting progress (show "X/Y posts this week")
 - Recent topics (avoid repetition within 7 days)
 - Current streak status
+- **`## Do-Next Directives`** — what the last measurement pass (report / analyze / A/B Adopt /
+  48h monitor) decided the next post should do differently, each with its evidence pointer.
+  Apply the ones that fit and name the applied directive in one line with the draft. Never edit
+  or clear the section (each source manages its own rows; rows older than 60 days expire).
 
 Read `skills/linkedin-studio/SKILL.md` for:
 - User's voice profile and tone preferences

commands/report.md

@@ -336,6 +336,40 @@ Delegate interpretation to the `analytics-interpreter` agent (report mode) — i
 3. **Test carousel format**: Your one carousel got 3x more shares than text posts. Create a carousel for your top-performing topic this week.
 ```
 
+## Step 7b: Persist them as do-next directives (the measurement→creation contract)
+
+A recommendation that only lands in chat dies there — the next post is drafted in a fresh
+session that never saw it. **Persist the 2–3 recommendations above as do-next directives**, so
+`/linkedin:post`, `/linkedin:quick`, `/linkedin:batch`, `/linkedin:create` and
+`/linkedin:newsletter` read them at their Step 0 and the next piece is actually shaped by this
+report. This step is not optional when a report produced recommendations.
+
+Each directive is an **imperative the next draft can act on** plus an **evidence pointer** —
+the measured number it came from, so the create surface can tell a directive backed by 12 posts
+from one backed by 2:
+
+```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: 'report',
+  directives: [
+    { directive: 'Plan 2 AI-agent posts this week', evidence: 'weekly 2026-W22: 3 AI posts avg 2400 impressions vs 1200 baseline (n=12)' },
+    { directive: 'Move Tuesday posts to 08:00', evidence: 'weekly 2026-W22: Tuesday -40% impressions' }
+  ]
+}));
+"
+```
+
+**Lifetime (do not hand-manage it):** `recordDoNext` replaces this source's own previous rows on
+every write and drops any row older than 60 days, so the section stays a live steering signal
+rather than a growing backlog. Directives from other sources (`ab-test`, `analyze`,
+`48h-monitor`) are left alone — never clear the section by hand.
+
+Then confirm in one line what was persisted, e.g. "2 do-next directives written to state — the
+next post/newsletter will read them at Step 0."
+
 ## Step 8: Offer Deep Dive Options
 
 After presenting the report, ask: