diff --git a/README.md b/README.md
index 3d7ba55..4249924 100644
--- a/README.md
+++ b/README.md
@@ -120,7 +120,11 @@ Build authentic LinkedIn authority through algorithmic understanding, strategic
 
 - **Long-form newsletter pipeline** — a multi-phase orchestrator (research → skeleton → prose → de-AI/voice scrub → fact-check → editorial and persona gates → visual assets → lock → delivery) with maintained edition state
 - **Adversarial review** — cold headless content, language, and fact reviewers re-check a frozen draft with no drafting-session context; `/linkedin:pivot` re-opens cleared gates on major rewrites
-- **Content engine** — Content Matrix (40+ ideas from one topic), voice training with drift detection, full ideation → publish → 48-hour monitoring → analytics
+- **Content engine** — Content Matrix (40+ ideas from one topic), voice training with drift detection, full ideation → publish → 48-hour monitoring
+- **Honest analytics** — CSV-import pipeline with weekly/monthly reports and anomaly alerts. Adds an
+  **optional, manually-entered `saves`** count (add a `Saves` column with the number read off native
+  LinkedIn analytics; never auto-tracked, `unknown` ≠ 0, excluded from the engagement rate). `dwell`
+  time stays **explicitly unmeasurable** — no fabricated metric or surface
 - **Growth and monetization** — phase-specific guidance from 0 to 10K+ followers; topic-relevance profile optimization aligned to LinkedIn's 2026 ranking model
 
 Key commands: `/linkedin:create` + `/linkedin:measure` (journey front-doors), `/linkedin:onboarding`, `/linkedin:post`, `/linkedin:newsletter`, `/linkedin:report`
diff --git a/plugins/linkedin-studio/CHANGELOG.md b/plugins/linkedin-studio/CHANGELOG.md
index e04658d..a225a27 100644
--- a/plugins/linkedin-studio/CHANGELOG.md
+++ b/plugins/linkedin-studio/CHANGELOG.md
@@ -25,6 +25,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Compatibility
 - **Minor / additive.** No command removed, renamed, or behavior-changed; the 27 existing commands, all 19 agents, and all state shapes are unchanged. The two new commands (`create`, `measure`) register when the plugin command set is rebuilt at session start — **reload required** to see them.
 
+### Added — within 4.1.0 (refinement sessions, no surface/count/version change)
+- **Manual per-post saves in analytics (Voyage S16).** Lifts the original v4.0.0 Non-Goal: `PostMetrics` gains an **optional** `saves` field, ingested when the user adds a `Saves` column to the CSV with the count read off native LinkedIn post analytics (count-only, ~Sept 2025; absent from the export, no self-serve API). The parser (`scripts/analytics/src/parsers/csv-parser.ts`) reads it when present; weekly/monthly summaries gain an optional `totalSaves`; the CLI (`import`/`report`) surfaces saves per-post and as a total. **Backward-compatible** — a missing column or blank cell leaves saves *unknown* (never coerced to 0), saves is **not** folded into `engagementRate` (which stays comparable to older imports), and saves-free data round-trips byte-identical. **Dwell stays explicitly unmeasurable** — no dwell field or surface was added. This refines the v4.0.0 "the plugin cannot read those signals" wording: the plugin still cannot *auto-track* saves, but it now ingests a *manually-entered* count. Built location-agnostically through the existing `getAnalyticsRoot()` seam so the planned data-dir migration (UI brief §9b/M0) relocates it in one place. New `RankableMetric` type fixes the trend/alert index access that the optional field would otherwise widen to `number | undefined`.
+- **Onboarding tool-grant fix (S16-pre).** `commands/onboarding.md` Phase 2 saves voice/user-profile files but its frontmatter omitted `Write`; added `Write` to `allowed-tools` (matching `first-post.md`). Closes a pre-existing tool-contract gap surfaced by the S15 review.
+
 ## [4.0.0] - 2026-05-30
 
 ### Summary
diff --git a/plugins/linkedin-studio/CLAUDE.md b/plugins/linkedin-studio/CLAUDE.md
index 9d89249..a29d74e 100644
--- a/plugins/linkedin-studio/CLAUDE.md
+++ b/plugins/linkedin-studio/CLAUDE.md
@@ -11,6 +11,17 @@ Full-spectrum LinkedIn content engine — short-form feed posts, carousels, vide
 - **Post queue:** `assets/drafts/queue.json` (managed by `hooks/scripts/queue-manager.mjs`)
 - **Analytics CLI:** `scripts/analytics/` (TypeScript, requires `tsx` and `npm install`)
 - **Analytics data:** `assets/analytics/` (gitignored)
+- **Analytics metrics (S16):** the parsed CSV columns plus an **optional, manually-entered** `saves` count.
+  Saves are count-only in native LinkedIn post analytics (~Sept 2025), absent from the CSV export, and
+  have no self-serve API — so the ingest path is the user adding a `Saves` column with the number they
+  read off LinkedIn. `parseOptionalCount()` parses it when present: blank / non-numeric / negative →
+  `undefined` (`unknown`, never 0), a genuine `0` is kept, and saves is **not** folded into
+  `engagementRate` (kept comparable to older imports). Surfaced per-post + as `totalSaves` in the
+  weekly/monthly reports; **never auto-tracked**.
+- **Unmeasured by design:** `dwell` time stays **explicitly unmeasurable** — internal to LinkedIn for
+  organic posts, no exportable count, no API; no dwell field or surface exists. The S16 analytics
+  extension routes all I/O through the existing `getAnalyticsRoot()` seam, so the planned per-user
+  data-dir migration (UI-brief §9b/M0) relocates the root in one place without reworking the schema.
 
 ## Hooks
 
diff --git a/plugins/linkedin-studio/README.md b/plugins/linkedin-studio/README.md
index a2386aa..b9a9f2b 100644
--- a/plugins/linkedin-studio/README.md
+++ b/plugins/linkedin-studio/README.md
@@ -437,9 +437,13 @@ profile**. As of 2026-05:
   Community Management API app + a verified organization + a Page). It is **not
   self-serve** for a solo personal profile, so the practical floor is the **CSV
   export** you drop into `/linkedin:import`. Per-post **saves** are visible in your
-  **native** LinkedIn post analytics (count-only, since ~Sept 2025), but there is
-  no self-serve API to pull them — this tool does not auto-track them; read them in
-  LinkedIn directly.
+  **native** LinkedIn post analytics (count-only, since ~Sept 2025) but are absent
+  from the CSV export, and there is no self-serve API to pull them. The tool does
+  **not** auto-track saves — but you can record them **manually**: add a `Saves`
+  column to the CSV with the counts you read off LinkedIn, and `/linkedin:import`
+  ingests them and surfaces them in the weekly/monthly reports (omit the column and
+  saves simply stays unknown — never counted as 0, never folded into engagement
+  rate).
 - **Auto-publish** — technically **possible** self-serve via the `w_member_social`
   scope, so this is **not** an API limitation. LinkedIn Studio **deliberately does
   not** post for you: the OAuth/token overhead plus LinkedIn's terms on automated
diff --git a/plugins/linkedin-studio/assets/analytics/README.md b/plugins/linkedin-studio/assets/analytics/README.md
index 4503998..16047b7 100644
--- a/plugins/linkedin-studio/assets/analytics/README.md
+++ b/plugins/linkedin-studio/assets/analytics/README.md
@@ -9,6 +9,25 @@ This directory contains imported analytics data from LinkedIn CSV exports.
 3. Save the CSV file to `exports/` directory
 4. Run `/linkedin:import` in Claude Code
 
+### Optional: add per-post saves (manual)
+
+LinkedIn's CSV export does **not** include saves, and there is no self-serve API
+to pull them — but the per-post save **count** is visible in your native post
+analytics (since ~Sept 2025). To track it, add a `Saves` column to the CSV and
+type the count you read off LinkedIn. The importer picks it up automatically when
+the column is present:
+
+```
+"Content","Date","Impressions","Reactions","Comments","Shares","Clicks","Saves"
+"My post...",2026-02-10,5000,100,30,15,200,42
+```
+
+A missing column — or a blank `Saves` cell — leaves saves **unknown** (never
+counted as 0), and saves is **not** folded into the engagement rate (which stays
+comparable to older imports). Saves is the strongest organic engagement signal,
+so the reports surface it as its own line. **Dwell time stays unmeasurable** —
+it is internal to LinkedIn for organic posts, with no count to transcribe.
+
 ## Directory Structure
 
 ```
@@ -50,6 +69,10 @@ Each file contains a batch of imported posts:
 }
 ```
 
+`metrics.saves` is **optional** — present only on posts where you supplied a
+`Saves` column value (see "Optional: add per-post saves" above). Posts without
+it omit the field entirely, so older imports round-trip unchanged.
+
 ### Weekly Reports (weekly-reports/*.json)
 
 Generated via `/linkedin:report`. Contains:
diff --git a/plugins/linkedin-studio/commands/onboarding.md b/plugins/linkedin-studio/commands/onboarding.md
index 39bc712..f262220 100644
--- a/plugins/linkedin-studio/commands/onboarding.md
+++ b/plugins/linkedin-studio/commands/onboarding.md
@@ -8,6 +8,7 @@ description: |
   "just installed", "how do I start", "walk me through", "linkedin onboarding".
 allowed-tools:
   - Read
+  - Write
   - Bash
   - AskUserQuestion
 ---
diff --git a/plugins/linkedin-studio/docs/remediation/review.md b/plugins/linkedin-studio/docs/remediation/review.md
index b64d3a0..b1ade77 100644
--- a/plugins/linkedin-studio/docs/remediation/review.md
+++ b/plugins/linkedin-studio/docs/remediation/review.md
@@ -1,13 +1,13 @@
 ---
 type: trekreview
 review_version: "1.0"
-task: "S15 — UX finish §6c on the final 29-command set: B1 onboarding inline-draft + B3 carousel full-deck clipboard (B2 router-tiering already delivered in S14)"
+task: "S16 — saves manual-entry surface (lifts the original Non-Goal): optional saves in the analytics data model (types + parser + weekly/monthly + CLI), built location-agnostic for M0 (option c); + S16-pre: close the deferred onboarding Write MAJOR"
 slug: remediation
 project_dir: docs/remediation/
 brief_path: docs/remediation/brief.md
-scope_sha_start: baca30f
-scope_sha_end: baca30f
-reviewed_files_count: 2
+scope_sha_start: 8c52bdb
+scope_sha_end: 8c52bdb
+reviewed_files_count: 15
 verdict: ALLOW
 mode: default
 effort: standard
@@ -15,129 +15,140 @@ profile: premium
 findings: []
 ---
 
-# Review — linkedin-studio S15 (UX finish §6c: B1 onboarding inline + B3 carousel full-deck)
+# Review — linkedin-studio S16 (saves manual-entry + S16-pre onboarding Write)
 
 ## Executive Summary
 
-**Verdict: ALLOW** for S15's delivered scope — 0 BLOCKER, 0 MAJOR, 0 MINOR, 0 SUGGESTION
-**attributable to the S15 diff**. Two independent reviewers (brief-conformance,
-code-correctness) ran COLD, without cross-feeding, on the as-delivered uncommitted
-working tree (HEAD `baca30f` + the S15 delta = two command files).
+**Verdict: ALLOW** for S16's delivered scope — 0 BLOCKER, 0 MAJOR, 0 MINOR, 0 SUGGESTION
+**open** against the diff. Two independent reviewers (brief-conformance, code-correctness)
+ran COLD, without cross-feeding, on the as-delivered uncommitted working tree (HEAD
+`8c52bdb` + the S16 delta).
 
-- **brief-conformance-reviewer:** 0 findings. B1 and B3 both trace to delivered code; the
-  binding §12b scope-guard is honored (no extensibility / provider-seam / progressive-
-  disclosure / detect-and-offer language entered the diff); no version/count drift
-  (19/29/25/6 intact, v4.1.0 untouched); no files outside the two in scope.
-- **code-correctness-reviewer:** 1 MAJOR — but **pre-existing and outside S15's scope**
-  (see Findings). The S15 changes themselves are correct: branches all terminate, the
-  state-updater + clipboard snippets mirror the established `first-post.md` pattern
-  exactly, `${CLAUDE_PLUGIN_ROOT}` is the established path var, and the larger carousel
-  clipboard payload introduces no new failure mode over the uniform plugin pattern.
+- **brief-conformance-reviewer:** **0 findings.** Every Success Criterion traces to delivered
+  code (SC1 saves in types + parser + import/report; SC2 saves in the actionable-signal output;
+  SC3 dwell still explicitly unmeasurable, no fabricated field; SC4 backward-compat + no-crash;
+  SC5 onboarding `Write`). Every Non-Goal honored: **M0 not built** (I/O still routes through the
+  unmodified `getAnalyticsRoot()` seam — option (c)), dwell not fabricated, surface counts
+  (29/19/6/9) and version (v4.1.0) unchanged, no not-mine file touched.
+- **code-correctness-reviewer:** **2 MAJOR — both in S16's own new code, both FIXED in-session**
+  (see Findings). Neither is pre-existing; both are lockstep misses in the saves parser delivered
+  this session, so per the operator rule ("in-session fix of the session's *own* misses =
+  completion") they were corrected here, not deferred.
 
-S15 = **B1 + B3 only**. B2 (router tiering) was already delivered in S14 (the finish-plan
-S15 body still lists B2, but the S14 amendment + the session-state label supersede it);
-its absence here is correct, not a gap.
+The M0 conflict (UI-brief §9b) was reconciled **before** building, by operator decision =
+**option (c): build now, location-agnostic**. The conformance reviewer independently confirmed
+the build honors it (no new hardcoded `assets/analytics` path; `storage.ts` untouched).
 
 ## Coverage
 
-Scope: HEAD `baca30f` (S14's commit) + the **uncommitted S15 working-tree delta**
-(annotated `[uncommitted]` — a brief-level contract; the brief's Assumptions allow
-uncommitted review). 2 files = the operator's own changes. The 3 untracked not-mine files
-(`docs/linkedin-studio-persona-brief.md`, `…-ui-brief.md`, `docs/voyage-build/progress.json`)
-are explicitly excluded from scope and from the commit. **No silent skips.**
+Scope: HEAD `8c52bdb` (S15's commit) + the **uncommitted S16 working-tree delta** (annotated
+`[uncommitted]` — a brief-level contract; the brief's Assumptions allow uncommitted review).
+The 3 untracked not-mine files (`docs/linkedin-studio-persona-brief.md`, `…-ui-brief.md`,
+`docs/voyage-build/progress.json`) are explicitly excluded from scope and from the commit.
+**No silent skips.**
 
 | Treatment | Count | Notes |
 |-----------|-------|-------|
-| `deep-review` | 0 | neither file is under `hooks/**` / `auth/**` / `crypto/**` / `**/security/**` |
-| `summary-only` | 2 | `commands/onboarding.md` (Phase 3 rewrite + 1 Phase-4 summary line), `commands/carousel.md` (Step 6 clipboard) |
+| `deep-review` | 0 | nothing under `hooks/**` / `auth/**` / `crypto/**` / `**/security/**` (analytics CLI + one command frontmatter) |
+| `summary-only` | 15 | analytics src (5) + analytics tests (3) + 2 fixtures + assets/analytics/README.md + README.md + CHANGELOG.md + commands/onboarding.md |
 | `skip` | 0 | no lockfiles / svg / generated / dist |
 
-**B1/B3 acceptance checks (orchestrator-run greps):**
-- B1: `grep -E "[Rr]un \`?/linkedin:first-post" commands/onboarding.md` → **0 dead-end strings**;
-  inline draft steps 3.1–3.5 present (topic → 3-line draft → QC → present+clipboard → state-update that sets `first_post_date`).
-- B3: full-deck assembly block at `carousel.md:191` precedes the `clipboard-helper.mjs` call at `:211`; payload contains slide text; caption-only `<CAROUSEL_CAPTION>` removed.
-
-**Cross-cutting execution criteria:** `scripts/test-runner.sh` → **74 passed / 0 failed /
-0 warnings**, exit 0. `node --test hooks/scripts/__tests__/*.test.mjs` → **98/98** (no hook
-logic changed). `ls commands/*.md` → **29** (no surface-count change; no version bump —
-consistent with S11–S13 precedent: within-scope refinements stay at the current version).
+**Execution criteria (orchestrator-run):**
+- `npx tsc --noEmit` (analytics) → **clean** (the new `RankableMetric` type closes the
+  `metrics[keyof]` widening the optional `saves` would otherwise introduce in `alerts.ts` + `cli.ts`).
+- `node --import tsx --test tests/*.test.ts` (analytics) → **116/116**.
+- `bash scripts/test-runner.sh` → **74 passed / 0 failed / 0 warnings**, exit 0.
+- `node --test hooks/scripts/__tests__/*.test.mjs` → **98/98** (no hook logic changed).
+- `ls commands/*.md` → **29** (no surface-count change; no version bump — within-v4.1.0 refinement, S11–S13 precedent).
+- **E2E:** import + weekly + monthly with a saves fixture surfaces saves without crashing;
+  saves-free fixtures round-trip byte-identical; explicit `"0"` → `0`, blank/`"n/a"` → undefined.
 
 ## Findings
 
-**0 findings attributable to the S15 diff.** One MAJOR was raised by code-correctness; on
-inspection it is **pre-existing and out of S15's scope**, and is routed to the next session
-per the finish-plan's locked constraint ("fix-in-next-session for any review finding") and
-the operator rule "ekte design-funn → neste sesjon" (in-session fix is reserved for the
-session's *own* lockstep misses). It is recorded below and propagated to STATE.md — **not
-dropped, not silenced.**
+**0 open findings.** Code-correctness raised 2 MAJOR; both are **S16's own** new code (not
+pre-existing — the saves parser is new this session), so both were **fixed in-session** as
+completion of the delivered work, then re-verified green. Recorded below — not dropped, not
+silenced.
 
-### [MAJOR — DEFERRED to next session] onboarding Phase 2 file-saves need `Write`, not in `allowed-tools`
+### [MAJOR — FIXED in-session] Non-numeric Saves cell silently coerced to 0 (`unknown != 0` contract violated)
 
-*Raised by code-correctness (`PLAN_EXECUTE_DRIFT`), `commands/onboarding.md:142` (and `:157`).*
+*Raised by code-correctness (`PLAN_EXECUTE_DRIFT`), `scripts/analytics/src/parsers/csv-parser.ts`.*
 
-`onboarding.md` frontmatter grants `allowed-tools: [Read, Bash, AskUserQuestion]` — no
-`Write`. **Phase 2** instructs "Save the responses to `assets/voice-samples/authentic-voice-samples.md`"
-(`:142`, with a REPLACE-vs-append rule) and "Save to `config/user-profile.local.md`" (`:157`).
-Persisting those profile files is a file write; the plugin's accepted pattern is the `Write`
-tool, and the sibling `first-post.md` (same voice-save) declares `Write` (`first-post.md:9-14`).
-With only Read/Bash/AskUserQuestion the agent cannot honor the two "Save to …" instructions
-(the only scripted Bash writes are the `node -e` state and clipboard snippets). **Real defect.**
+The original guard `if (savesRaw && savesRaw.trim() !== "")` then `metrics.saves =
+parseMetric(savesRaw)` gated only on emptiness; `parseMetric` ends in `parseFloat(...) || 0`
++ `Math.max(0, …)`, so a non-empty non-numeric cell (`"n/a"`, `"~40"`) or a negative value
+flattened to **0** — surfaced as a confident "top engagement signal". This contradicts the
+`unknown != 0` contract stated in the same diff (`types.ts` saves NOTE + the parser comment).
 
-**Why DEFERRED, not fixed in S15:**
-- **Pre-existing.** The gap exists identically at HEAD `baca30f` *before* the S15 edits —
-  `git show HEAD:…/onboarding.md` shows `allowed-tools: [Read, Bash, AskUserQuestion]` already.
-- **Out of S15's diff scope.** The S15 hunks are `@@ -166` (Phase 3) and `@@ -201` (Phase 4
-  summary line); lines 142/157 (Phase 2) are untouched. S15's *own* inline steps (3.1–3.5)
-  stay within Read/Bash/AskUserQuestion and need no `Write` (3.4 clipboard = Bash, 3.5
-  state-update = Bash `node -e`).
-- **Not introduced or worsened by S15.** The reviewer's "now owns the save flow" framing does
-  not hold on inspection: Phase 2's saves are reached identically before and after S15, and
-  the removed `/linkedin:first-post` hand-off was about *post creation* (Phase 3), not Phase 2's
-  profile saves. S15 added no save-flow that needs `Write`.
-- **Operator governance.** Fixing a Phase-2 tool-contract bug inside an S15 (B1+B3) push would
-  be exactly the "rydd opp i pre-existing som del av en bugfix" / scope-expansion the operator's
-  rules forbid; the documented default for out-of-scope items is **defer** (record, route to
-  next session) — not silently fix, not silently drop.
+**Fix (this session):** added a dedicated `parseOptionalCount()` helper returning
+`number | undefined` — blank / non-numeric / negative → `undefined`; a genuine number
+(including explicit `"0"`) → that number; reuses the EU/US separator normalization. The saves
+parse now routes through it. The blank-cell path is unchanged; the garbage/negative path now
+honors the contract.
 
-**Recommended action (next session):** add `Write` to `onboarding.md`'s `allowed-tools`
-(matching `first-post.md:9-14`), or replace the two "Save to …" instructions with a `node -e`
-Bash write so the steps stay within the declared Bash grant. Natural home: fold into S16 (which
-already opens the onboarding/saves surface) or the S17 triage pass.
+### [MAJOR — FIXED in-session] No test for the explicit-`0` vs non-numeric Saves boundary
+
+*Raised by code-correctness (`MISSING_TEST`), `scripts/analytics/tests/csv-parser.test.ts`.*
+
+The original saves tests covered `42`, blank→undefined, no-column→undefined, and
+engagement-exclusion — but not the exact boundary where the parser's behavior diverged from
+its intent: a literal `"0"` (genuine zero) vs a non-numeric cell (unknown).
+
+**Fix (this session):** added `tests/fixtures/saves-edge-export.csv` (a `"0"` row + an `"n/a"`
+row) and two cases pinning `saves === 0` for `"0"` and `saves === undefined` for non-numeric.
+The non-numeric case was **red** before the parser fix above and **green** after (TDD), so the
+contract boundary is now regression-guarded.
+
+### [MAJOR — CLOSED] (carried from S15) onboarding Phase 2 file-saves need `Write`
+
+*S15's deferred finding.* `commands/onboarding.md` Phase 2 saves voice/user-profile files but
+the frontmatter omitted `Write`. **Closed in S16-pre:** `Write` added to `allowed-tools`
+(matching `first-post.md:9-14`).
 
 ## Remediation Summary
 
-**Gate: ALLOW** for S15's delivered scope (B1 + B3). Both reviewers confirm the two changes
-are conformant (brief) and correct (code); the binding §12b scope-guard held; no version/count
-drift; the inline first-post flow and the full-deck clipboard both function within the existing
-tool grants and established patterns. The single MAJOR is a **pre-existing, out-of-scope**
-tool-contract gap in Phase 2 that S15 neither introduced nor worsened; per the operator's
-explicit "ekte design-funn → neste sesjon" rule it is **deferred and recorded**, not fixed
-as scope creep. This is a genuine ALLOW of the scoped delivery — **not** a WARN-override (there
-is no open finding *against* S15's diff).
+**Gate: ALLOW** for S16's delivered scope. brief-conformance is clean; code-correctness's two
+MAJORs were S16's own lockstep misses (not pre-existing design findings), so both were fixed
+in-session and re-verified (tsc clean, analytics 116/116, lint 74/0/0, hooks 98/98) — this is a
+genuine ALLOW with **no open finding**, not a WARN-override. The M0 conflict was reconciled
+before building (option c, location-agnostic); the conformance reviewer confirmed M0 was not
+built and the data-dir seam (`getAnalyticsRoot()`) is untouched, so the planned migration
+relocates the root in one place. The S15-deferred onboarding `Write` MAJOR is closed.
 
-Per Handover 6, this `review.md` is consumable by `/trekplan --brief …`. With an ALLOW verdict
-on the delivered scope and the one deferred finding routed forward, S15 may commit + push, and
-the finish-plan continues at S16 (with the deferred `Write` gap reconciled there or in S17).
+Per Handover 6, this `review.md` is consumable by `/trekplan --brief …`. ALLOW → S16 commits +
+pushes (own files only); the finish-plan continues at **S17** (C13–C46 triage).
 
 ```json
 {
   "verdict": "ALLOW",
-  "verdict_scope": "S15 delivered changes (B1 + B3); 2 files",
-  "scope": { "sha_start": "baca30f", "sha_end": "baca30f", "reviewed_files_count": 2, "uncommitted_delta": true },
+  "verdict_scope": "S16 delivered changes (saves manual-entry + S16-pre onboarding Write); 15 files",
+  "scope": { "sha_start": "8c52bdb", "sha_end": "8c52bdb", "reviewed_files_count": 15, "uncommitted_delta": true },
   "counts": { "BLOCKER": 0, "MAJOR": 0, "MINOR": 0, "SUGGESTION": 0 },
   "findings": [],
-  "deferred_findings": [
+  "fixed_in_session": [
     {
       "severity": "MAJOR",
-      "title": "onboarding Phase 2 file-saves need Write, not in allowed-tools",
-      "file": "commands/onboarding.md",
-      "line": 142,
+      "title": "Non-numeric Saves cell silently coerced to 0 (unknown != 0 contract)",
+      "file": "scripts/analytics/src/parsers/csv-parser.ts",
       "rule_key": "PLAN_EXECUTE_DRIFT",
-      "status": "pre-existing, out of S15 scope (Phase 2; S15 diff = Phase 3 + Phase 4 line)",
-      "routed_to": "next session (S16 onboarding/saves surface, or S17 triage)",
-      "recommended_action": "add Write to onboarding.md allowed-tools (match first-post.md:9-14), or replace the two 'Save to ...' instructions with a node -e Bash write"
+      "resolution": "added parseOptionalCount() returning number|undefined (blank/non-numeric/negative -> undefined; genuine 0 kept); saves parse routes through it"
+    },
+    {
+      "severity": "MAJOR",
+      "title": "No test for explicit-0 vs non-numeric Saves boundary",
+      "file": "scripts/analytics/tests/csv-parser.test.ts",
+      "rule_key": "MISSING_TEST",
+      "resolution": "added saves-edge-export.csv fixture + 2 cases (0 -> 0, n/a -> undefined); TDD red-before/green-after the parser fix"
+    },
+    {
+      "severity": "MAJOR",
+      "title": "onboarding Phase 2 file-saves need Write (carried from S15)",
+      "file": "commands/onboarding.md",
+      "rule_key": "PLAN_EXECUTE_DRIFT",
+      "resolution": "S16-pre: added Write to allowed-tools, matching first-post.md"
     }
   ],
+  "deferred_findings": [],
   "dropped_findings": []
 }
 ```
diff --git a/plugins/linkedin-studio/scripts/analytics/src/cli.ts b/plugins/linkedin-studio/scripts/analytics/src/cli.ts
index c6c1e1b..2024815 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/cli.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/cli.ts
@@ -12,7 +12,7 @@ import { generateHeatmap } from "./reports/heatmap.js";
 import { generateMonthlyReport } from "./reports/monthly.js";
 import { join } from "node:path";
 import { existsSync } from "node:fs";
-import type { PostMetrics } from "./models/types.js";
+import type { RankableMetric } from "./models/types.js";
 
 const args = process.argv.slice(2);
 const command = args[0];
@@ -22,6 +22,14 @@ function parseOption(args: string[], flag: string): string | undefined {
   return idx !== -1 && idx + 1 < args.length ? args[idx + 1] : undefined;
 }
 
+/**
+ * Per-post saves suffix for report lines. Empty string when the post carries no
+ * manual saves data, so saves-free output stays identical to the pre-saves CLI.
+ */
+function savesSuffix(saves?: number): string {
+  return saves !== undefined ? ` | ${saves.toLocaleString()} saves` : "";
+}
+
 function printUsage() {
   console.log(`
 LinkedIn Analytics CLI
@@ -75,6 +83,13 @@ async function handleImport(root: string, args: string[]) {
     console.log(`Batch ID:        ${batch.batchId}`);
     console.log(`Saved to:        posts/${savedFilename}`);
 
+    // Surface manually-entered saves when the CSV carried a Saves column.
+    const savesPosts = batch.posts.filter((p) => p.metrics.saves !== undefined);
+    if (savesPosts.length > 0) {
+      const totalSaves = savesPosts.reduce((sum, p) => sum + (p.metrics.saves ?? 0), 0);
+      console.log(`Saves entered:   ${totalSaves.toLocaleString()} across ${savesPosts.length} post(s) (manual)`);
+    }
+
     // Run alert detection on imported posts
     const alerts = detectAlerts(batch.posts, "impressions");
 
@@ -126,6 +141,9 @@ async function handleReport(root: string, args: string[]) {
     console.log(`Total comments:      ${report.summary.totalComments.toLocaleString()}`);
     console.log(`Total shares:        ${report.summary.totalShares.toLocaleString()}`);
     console.log(`Total clicks:        ${report.summary.totalClicks.toLocaleString()}`);
+    if (report.summary.totalSaves !== undefined) {
+      console.log(`Total saves:         ${report.summary.totalSaves.toLocaleString()} (manual entry — top engagement signal)`);
+    }
     console.log(`Avg engagement:      ${report.summary.avgEngagementRate.toFixed(2)}%`);
     console.log(`Avg impressions:     ${Math.round(report.summary.avgImpressionsPerPost).toLocaleString()} per post`);
     console.log();
@@ -136,7 +154,7 @@ async function handleReport(root: string, args: string[]) {
       for (const post of report.topPerformers.slice(0, 5)) {
         const title = post.title.length > 50 ? post.title.substring(0, 47) + "..." : post.title;
         console.log(`• ${title}`);
-        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% engagement | ${post.publishedDate}`);
+        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% engagement${savesSuffix(post.metrics.saves)} | ${post.publishedDate}`);
       }
       console.log();
     }
@@ -147,7 +165,7 @@ async function handleReport(root: string, args: string[]) {
       for (const post of report.underperformers.slice(0, 3)) {
         const title = post.title.length > 50 ? post.title.substring(0, 47) + "..." : post.title;
         console.log(`• ${title}`);
-        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% engagement | ${post.publishedDate}`);
+        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% engagement${savesSuffix(post.metrics.saves)} | ${post.publishedDate}`);
       }
       console.log();
     }
@@ -177,10 +195,11 @@ async function handleReport(root: string, args: string[]) {
 }
 
 /**
- * Type guard to check if a string is a valid PostMetrics key
+ * Type guard to check if a string is a rankable (always-numeric) metric key.
+ * Excludes the optional, manually-entered `saves` — it is not a trend metric.
  */
-function isPostMetric(value: string): value is keyof PostMetrics {
-  const validMetrics: (keyof PostMetrics)[] = [
+function isPostMetric(value: string): value is RankableMetric {
+  const validMetrics: RankableMetric[] = [
     "impressions",
     "reactions",
     "comments",
@@ -188,7 +207,7 @@ function isPostMetric(value: string): value is keyof PostMetrics {
     "clicks",
     "engagementRate",
   ];
-  return validMetrics.includes(value as keyof PostMetrics);
+  return validMetrics.includes(value as RankableMetric);
 }
 
 async function handleTrends(root: string, args: string[]) {
@@ -203,7 +222,7 @@ async function handleTrends(root: string, args: string[]) {
   }
 
   if (!isPostMetric(metricOption)) {
-    const validMetrics: (keyof PostMetrics)[] = [
+    const validMetrics: RankableMetric[] = [
       "impressions",
       "reactions",
       "comments",
@@ -320,6 +339,9 @@ async function handleMonthlyReport(root: string, month: string) {
     console.log(`Comments:            ${s.totalComments.toLocaleString()}`);
     console.log(`Shares:              ${s.totalShares.toLocaleString()}`);
     console.log(`Clicks:              ${s.totalClicks.toLocaleString()}`);
+    if (s.totalSaves !== undefined) {
+      console.log(`Saves:               ${s.totalSaves.toLocaleString()} (manual entry — top engagement signal)`);
+    }
     console.log();
 
     if (report.byWeek.length > 0) {
@@ -337,7 +359,7 @@ async function handleMonthlyReport(root: string, month: string) {
       for (const post of report.topPerformers.slice(0, 5)) {
         const title = post.title.length > 50 ? post.title.substring(0, 47) + "..." : post.title;
         console.log(`• ${title}`);
-        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% eng | ${post.publishedDate}`);
+        console.log(`  ${post.metrics.impressions.toLocaleString()} impressions | ${post.metrics.engagementRate.toFixed(2)}% eng${savesSuffix(post.metrics.saves)} | ${post.publishedDate}`);
       }
       console.log();
     }
diff --git a/plugins/linkedin-studio/scripts/analytics/src/models/types.ts b/plugins/linkedin-studio/scripts/analytics/src/models/types.ts
index d85ba5c..c9ed27f 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/models/types.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/models/types.ts
@@ -14,11 +14,18 @@ export interface PostMetrics {
   shares: number;
   clicks: number;
   engagementRate: number; // (reactions+comments+shares+clicks)/impressions * 100
-  // NOTE: `saves` and `dwell` are intentionally absent. They are NOT in the
-  // LinkedIn analytics CSV export this tool parses, and there is no self-serve API
-  // to pull them. Saves are visible (count-only) in the native post analytics UI
-  // (~Sept 2025 onward) — read them there; dwell is internal to LinkedIn for
-  // organic posts. Do not add these fields without a real ingest source.
+  // `saves` is OPTIONAL and manually entered. LinkedIn's CSV export does NOT
+  // include it and there is no self-serve API to pull it — but the count IS
+  // visible in the native post analytics UI (~Sept 2025 onward). The ingest
+  // path is the user adding a `Saves` column to the CSV they read off it; the
+  // parser picks it up when present (see csv-parser.ts). When the column or a
+  // cell is absent, `saves` stays undefined — "unknown", never coerced to 0.
+  // It is deliberately NOT folded into engagementRate (which stays comparable
+  // to historical, saves-free data) — saves is surfaced as its own signal.
+  saves?: number;
+  // NOTE: `dwell` remains absent and unmeasurable. Dwell time is internal to
+  // LinkedIn for organic posts — not exportable, no UI count to transcribe, no
+  // API. Do not fabricate a dwell field or surface.
 }
 
 export interface AnalyticsBatch {
@@ -40,6 +47,7 @@ export interface WeeklyReport {
     totalComments: number;
     totalShares: number;
     totalClicks: number;
+    totalSaves?: number; // optional — present only when ≥1 post carries manual saves data
     avgEngagementRate: number;
     avgImpressionsPerPost: number;
   };
@@ -59,6 +67,21 @@ export interface WeeklyReport {
 
 export type TrendDirection = "up" | "down" | "stable";
 
+/**
+ * Metric keys that are always present and numeric — safe for trend/alert ranking
+ * and `metrics[key]` index access. Excludes the optional, manually-entered
+ * `saves`, which is sparse and would type as `number | undefined` under index
+ * access (and is not a rankable trend metric). This is the runtime whitelist the
+ * CLI and alert engine have always used.
+ */
+export type RankableMetric =
+  | "impressions"
+  | "reactions"
+  | "comments"
+  | "shares"
+  | "clicks"
+  | "engagementRate";
+
 export interface Alert {
   type: "spike" | "drop" | "milestone";
   severity: "info" | "warning" | "critical";
@@ -98,6 +121,7 @@ export interface MonthlyReport {
     totalComments: number;
     totalShares: number;
     totalClicks: number;
+    totalSaves?: number; // optional — present only when ≥1 post carries manual saves data
     avgEngagementRate: number;
     avgImpressionsPerPost: number;
   };
diff --git a/plugins/linkedin-studio/scripts/analytics/src/parsers/csv-parser.ts b/plugins/linkedin-studio/scripts/analytics/src/parsers/csv-parser.ts
index 9edbbd5..229b21c 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/parsers/csv-parser.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/parsers/csv-parser.ts
@@ -57,6 +57,33 @@ function parseMetric(value: string): number {
   return Math.max(0, parsed);
 }
 
+/**
+ * Parse an OPTIONAL manually-entered count (saves). Unlike parseMetric — which
+ * coerces blanks, garbage, and negatives to 0 — this preserves the "unknown vs
+ * zero" distinction the saves contract requires:
+ *   - blank / absent          → undefined ("unknown", never 0)
+ *   - non-numeric ("n/a", …)  → undefined ("unknown", never 0)
+ *   - negative                → undefined (not a real save count)
+ *   - a genuine number ("0")  → that number (an explicit 0 is a real reading)
+ * Reuses the same EU/US thousand-separator normalization as parseMetric so a
+ * "1.234"/"1,234" Saves cell parses consistently with the other columns.
+ */
+function parseOptionalCount(value: string): number | undefined {
+  if (!value) return undefined;
+  const cleaned = value.replace(/"/g, "").trim();
+  if (cleaned === "") return undefined;
+
+  const lastComma = cleaned.lastIndexOf(",");
+  const lastDot = cleaned.lastIndexOf(".");
+  const normalized = lastComma > lastDot
+    ? cleaned.replace(/,/g, "")
+    : cleaned.replace(/\./g, "").replace(/,/g, ".");
+
+  const parsed = Number(normalized);
+  if (!Number.isFinite(parsed) || parsed < 0) return undefined;
+  return parsed;
+}
+
 /**
  * Normalizes date to YYYY-MM-DD format
  * Handles: DD.MM.YYYY, MM/DD/YYYY, YYYY-MM-DD
@@ -173,7 +200,8 @@ export function parseLinkedInCSV(
       const shares = parseMetric(findColumn(record, ["share", "repost"]));
       const clicks = parseMetric(findColumn(record, ["click"]));
 
-      // Calculate engagement rate
+      // Calculate engagement rate — saves is deliberately NOT in the numerator,
+      // so this stays comparable to historical, saves-free imports.
       const totalEngagement = reactions + comments + shares + clicks;
       const engagementRate = impressions > 0
         ? (totalEngagement / impressions) * 100
@@ -188,6 +216,15 @@ export function parseLinkedInCSV(
         engagementRate,
       };
 
+      // Optional manual-entry saves: only when the user augmented this CSV with a
+      // Saves column (read off native LinkedIn analytics, ~Sept 2025+). A missing
+      // column, a blank cell, or a non-numeric/negative cell stays undefined —
+      // "unknown", never coerced to 0; a genuine 0 is kept as 0.
+      const saves = parseOptionalCount(findColumn(record, ["saves", "bookmark"]));
+      if (saves !== undefined) {
+        metrics.saves = saves;
+      }
+
       return {
         id: generatePostId(title, date),
         title,
diff --git a/plugins/linkedin-studio/scripts/analytics/src/reports/monthly.ts b/plugins/linkedin-studio/scripts/analytics/src/reports/monthly.ts
index 571abdd..db9f965 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/reports/monthly.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/reports/monthly.ts
@@ -28,6 +28,12 @@ export function generateMonthlyReport(root: string, month: string): MonthlyRepor
   const totalComments = monthPosts.reduce((s, p) => s + p.metrics.comments, 0);
   const totalShares = monthPosts.reduce((s, p) => s + p.metrics.shares, 0);
   const totalClicks = monthPosts.reduce((s, p) => s + p.metrics.clicks, 0);
+  // Optional saves: present only when ≥1 post carries manual saves data —
+  // keeps saves-free months byte-identical to pre-saves output (backward-compat).
+  const savesPosts = monthPosts.filter(p => p.metrics.saves !== undefined);
+  const totalSaves = savesPosts.length > 0
+    ? savesPosts.reduce((s, p) => s + (p.metrics.saves ?? 0), 0)
+    : undefined;
   const avgEngagementRate = totalPosts > 0
     ? parseFloat(mean(monthPosts.map(p => p.metrics.engagementRate)).toFixed(2))
     : 0;
@@ -101,6 +107,7 @@ export function generateMonthlyReport(root: string, month: string): MonthlyRepor
       totalComments,
       totalShares,
       totalClicks,
+      ...(totalSaves !== undefined ? { totalSaves } : {}),
       avgEngagementRate,
       avgImpressionsPerPost,
     },
diff --git a/plugins/linkedin-studio/scripts/analytics/src/reports/weekly.ts b/plugins/linkedin-studio/scripts/analytics/src/reports/weekly.ts
index f1cc857..52c9901 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/reports/weekly.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/reports/weekly.ts
@@ -154,12 +154,23 @@ export function generateWeeklyReport(analyticsRoot: string, week?: string): Week
   }
 
   // Calculate summary metrics
+  let totalSaves = 0;
+  let sawSaves = false;
   for (const post of weekPosts) {
     report.summary.totalImpressions += post.metrics.impressions;
     report.summary.totalReactions += post.metrics.reactions;
     report.summary.totalComments += post.metrics.comments;
     report.summary.totalShares += post.metrics.shares;
     report.summary.totalClicks += post.metrics.clicks;
+    if (post.metrics.saves !== undefined) {
+      totalSaves += post.metrics.saves;
+      sawSaves = true;
+    }
+  }
+  // Only surface saves when at least one post carried it — keeps saves-free
+  // reports byte-identical to pre-saves output (backward-compat).
+  if (sawSaves) {
+    report.summary.totalSaves = totalSaves;
   }
 
   // Calculate averages
diff --git a/plugins/linkedin-studio/scripts/analytics/src/utils/alerts.ts b/plugins/linkedin-studio/scripts/analytics/src/utils/alerts.ts
index a845713..6b2aa29 100644
--- a/plugins/linkedin-studio/scripts/analytics/src/utils/alerts.ts
+++ b/plugins/linkedin-studio/scripts/analytics/src/utils/alerts.ts
@@ -1,7 +1,7 @@
 import type {
   PostAnalytics,
   Alert,
-  PostMetrics,
+  RankableMetric,
 } from "../models/types.js";
 import { ALERT_THRESHOLDS } from "../models/types.js";
 import {
@@ -17,7 +17,7 @@ import {
  */
 export function detectAlerts(
   posts: PostAnalytics[],
-  metricKey: keyof PostMetrics = "impressions"
+  metricKey: RankableMetric = "impressions"
 ): Alert[] {
   if (posts.length === 0) return [];
 
diff --git a/plugins/linkedin-studio/scripts/analytics/tests/csv-parser.test.ts b/plugins/linkedin-studio/scripts/analytics/tests/csv-parser.test.ts
index 318e90d..efc8833 100644
--- a/plugins/linkedin-studio/scripts/analytics/tests/csv-parser.test.ts
+++ b/plugins/linkedin-studio/scripts/analytics/tests/csv-parser.test.ts
@@ -122,3 +122,72 @@ describe("CSV Parser", () => {
     assert.equal(batch.dateRange.to, "2026-01-28", "Date range to should be latest date");
   });
 });
+
+describe("Saves (manual-entry, optional)", () => {
+  it("should parse a Saves column when the user augments the CSV with it", () => {
+    const filePath = join(fixturesDir, "saves-export.csv");
+    const batch = parseLinkedInCSV(filePath, "saves-export.csv");
+
+    assert.equal(batch.postCount, 2, "Should have 2 posts");
+
+    // Row 1 carries a saves count read from native LinkedIn analytics.
+    assert.equal(batch.posts[0].metrics.saves, 42, "Should parse the Saves cell value");
+  });
+
+  it("should leave saves undefined when the Saves cell is blank (unknown != zero)", () => {
+    const filePath = join(fixturesDir, "saves-export.csv");
+    const batch = parseLinkedInCSV(filePath, "saves-export.csv");
+
+    // Row 2's Saves cell is empty — saves is unknown, NOT zero.
+    assert.equal(
+      batch.posts[1].metrics.saves,
+      undefined,
+      "Blank Saves cell must stay undefined, never coerced to 0"
+    );
+  });
+
+  it("should leave saves undefined for a standard export with no Saves column (backward-compat)", () => {
+    const filePath = join(fixturesDir, "sample-export.csv");
+    const batch = parseLinkedInCSV(filePath, "sample-export.csv");
+
+    for (const post of batch.posts) {
+      assert.equal(
+        post.metrics.saves,
+        undefined,
+        "Existing CSV exports without a Saves column must round-trip unchanged"
+      );
+    }
+  });
+
+  it("should NOT fold saves into engagementRate (kept comparable to historical data)", () => {
+    const filePath = join(fixturesDir, "saves-export.csv");
+    const batch = parseLinkedInCSV(filePath, "saves-export.csv");
+
+    // Row 1: (100+30+15+200)/5000 * 100 = 6.9 — saves (42) must NOT be in the numerator.
+    const expectedRate = ((100 + 30 + 15 + 200) / 5000) * 100;
+    assert.ok(
+      Math.abs(batch.posts[0].metrics.engagementRate - expectedRate) < 0.01,
+      `engagementRate should exclude saves (~${expectedRate}), got ${batch.posts[0].metrics.engagementRate}`
+    );
+  });
+
+  it("should treat an explicit '0' Saves cell as a genuine zero (not undefined)", () => {
+    const filePath = join(fixturesDir, "saves-edge-export.csv");
+    const batch = parseLinkedInCSV(filePath, "saves-edge-export.csv");
+
+    // A literal 0 in the Saves column is a real reading — zero saves, not unknown.
+    assert.equal(batch.posts[0].metrics.saves, 0, "Explicit '0' must stay 0, not collapse to undefined");
+  });
+
+  it("should leave saves undefined for a non-numeric Saves cell (unknown, never coerced to 0)", () => {
+    const filePath = join(fixturesDir, "saves-edge-export.csv");
+    const batch = parseLinkedInCSV(filePath, "saves-edge-export.csv");
+
+    // "n/a" is not a count — saves stays unknown, NOT silently flattened to 0.
+    assert.equal(
+      batch.posts[1].metrics.saves,
+      undefined,
+      "Non-numeric Saves cell must stay undefined — never coerced to 0"
+    );
+  });
+});
diff --git a/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-edge-export.csv b/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-edge-export.csv
new file mode 100644
index 0000000..68b514d
--- /dev/null
+++ b/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-edge-export.csv
@@ -0,0 +1,3 @@
+"Content","Date","Impressions","Reactions","Comments","Shares","Clicks","Saves"
+"Explicit zero saves — a real reading of zero, must stay 0 not undefined...",2026-02-12,4000,80,25,10,150,0
+"Non-numeric saves cell — the user jotted a note, not a count; stays unknown...",2026-02-11,3500,70,22,9,130,n/a
diff --git a/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-export.csv b/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-export.csv
new file mode 100644
index 0000000..243f80c
--- /dev/null
+++ b/plugins/linkedin-studio/scripts/analytics/tests/fixtures/saves-export.csv
@@ -0,0 +1,3 @@
+"Content","Date","Impressions","Reactions","Comments","Shares","Clicks","Saves"
+"A save-worthy framework post the user augmented with the native saves count...",2026-02-10,5000,100,30,15,200,42
+"A post where the user left the Saves cell blank — unknown, not zero...",2026-02-09,3000,60,20,8,120,
diff --git a/plugins/linkedin-studio/scripts/analytics/tests/monthly.test.ts b/plugins/linkedin-studio/scripts/analytics/tests/monthly.test.ts
index 3ca5f53..9f2d107 100644
--- a/plugins/linkedin-studio/scripts/analytics/tests/monthly.test.ts
+++ b/plugins/linkedin-studio/scripts/analytics/tests/monthly.test.ts
@@ -83,6 +83,27 @@ describe("generateMonthlyReport", () => {
     assert.equal(report.summary.avgImpressionsPerPost, 2000);
   });
 
+  test("sums saves into totalSaves when posts carry manual saves data", () => {
+    const withSaves = (p: PostAnalytics, saves: number): PostAnalytics => ({
+      ...p,
+      metrics: { ...p.metrics, saves },
+    });
+    const posts: PostAnalytics[] = [
+      withSaves(createPost("2026-03-03", 1000, 3.0), 8),
+      createPost("2026-03-05", 2000, 4.0), // no saves — partial coverage
+      withSaves(createPost("2026-03-10", 1500, 3.5), 13),
+    ];
+    const root = setupTestRoot(posts);
+    const report = generateMonthlyReport(root, "2026-03");
+    assert.equal(report.summary.totalSaves, 21, "totalSaves should sum 8 + 13");
+  });
+
+  test("leaves totalSaves undefined for saves-free months (backward-compat)", () => {
+    const root = setupTestRoot(marchPosts);
+    const report = generateMonthlyReport(root, "2026-03");
+    assert.equal(report.summary.totalSaves, undefined);
+  });
+
   test("generates weekly breakdown within month", () => {
     const root = setupTestRoot(marchPosts);
     const report = generateMonthlyReport(root, "2026-03");
diff --git a/plugins/linkedin-studio/scripts/analytics/tests/weekly.test.ts b/plugins/linkedin-studio/scripts/analytics/tests/weekly.test.ts
index 1551a86..dbb5d94 100644
--- a/plugins/linkedin-studio/scripts/analytics/tests/weekly.test.ts
+++ b/plugins/linkedin-studio/scripts/analytics/tests/weekly.test.ts
@@ -275,6 +275,50 @@ describe("weekly", () => {
       assert.ok(Math.abs(report.summary.avgEngagementRate - 8.49) < 0.01);
     });
 
+    test("should sum saves into totalSaves when posts carry manual saves data", () => {
+      tempDir = setupTempDir();
+
+      const posts: PostAnalytics[] = [
+        createTestPost({
+          id: "saved1",
+          publishedDate: "2026-01-12", // 2026-W03
+          metrics: { impressions: 1000, reactions: 50, comments: 10, shares: 5, clicks: 20, engagementRate: 8.5, saves: 12 },
+        }),
+        createTestPost({
+          id: "saved2",
+          publishedDate: "2026-01-13", // 2026-W03
+          // No saves on this one — partial coverage must still sum what exists.
+          metrics: { impressions: 2000, reactions: 100, comments: 20, shares: 10, clicks: 40, engagementRate: 8.5 },
+        }),
+        createTestPost({
+          id: "saved3",
+          publishedDate: "2026-01-14", // 2026-W03
+          metrics: { impressions: 1500, reactions: 75, comments: 15, shares: 7, clicks: 30, engagementRate: 8.47, saves: 30 },
+        }),
+      ];
+
+      saveBatch(tempDir, createTestBatch({ dateRange: { from: "2026-01-12", to: "2026-01-14" }, posts }));
+
+      const report = generateWeeklyReport(tempDir, "2026-W03");
+
+      assert.equal(report.summary.totalSaves, 42, "totalSaves should sum the posts that carry saves (12 + 30)");
+    });
+
+    test("should leave totalSaves undefined when no post carries saves (backward-compat)", () => {
+      tempDir = setupTempDir();
+
+      const posts: PostAnalytics[] = [
+        createTestPost({ id: "nosave1", publishedDate: "2026-01-12" }),
+        createTestPost({ id: "nosave2", publishedDate: "2026-01-13" }),
+      ];
+
+      saveBatch(tempDir, createTestBatch({ dateRange: { from: "2026-01-12", to: "2026-01-13" }, posts }));
+
+      const report = generateWeeklyReport(tempDir, "2026-W03");
+
+      assert.equal(report.summary.totalSaves, undefined, "Saves-free data must not introduce a totalSaves field");
+    });
+
     test("should identify top performers and underperformers", () => {
       tempDir = setupTempDir();