diff --git a/bin/gstack-global-discover b/bin/gstack-global-discover
index ebffeeb9e5..8eb63f4b84 100755
Binary files a/bin/gstack-global-discover and b/bin/gstack-global-discover differ
diff --git a/gstack-upgrade/SKILL.md b/gstack-upgrade/SKILL.md
index f97f11fb74..9e01fcdfc5 100644
--- a/gstack-upgrade/SKILL.md
+++ b/gstack-upgrade/SKILL.md
@@ -136,7 +136,31 @@ cd "$INSTALL_DIR" && ./setup
 rm -rf "$INSTALL_DIR.bak" "$TMP_DIR"
 ```
 
-### Step 4.5: Sync local vendored copy
+### Step 4.5: Apply user overlay
+
+After `./setup` completes, check if the user has customizations that should be restored.
+
+```bash
+_OVERLAY_DIR="$HOME/.gstack/user-overlay"
+if [ -d "$_OVERLAY_DIR" ]; then
+  _SKILL_NAME=$(basename "$INSTALL_DIR")
+  if [ -d "$_OVERLAY_DIR/$_SKILL_NAME" ]; then
+    cp -Rf "$_OVERLAY_DIR/$_SKILL_NAME/." "$INSTALL_DIR/"
+    echo "USER_OVERLAY: Applied overlay for $_SKILL_NAME"
+  fi
+fi
+```
+
+If overlay files exist, they are copied over the freshly-upgraded skill directory. This preserves user customizations (edited SKILL.md files, added templates, extra scripts) across gstack updates.
+
+Users create overlays by copying files to `~/.gstack/user-overlay/<skill-name>/`. For example, to persist edits to `plan-eng-review`:
+
+```bash
+mkdir -p ~/.gstack/user-overlay/plan-eng-review
+cp ~/.claude/skills/gstack/plan-eng-review/SKILL.md.tmpl ~/.gstack/user-overlay/plan-eng-review/
+```
+
+### Step 4.6: Sync local vendored copy
 
 Use the install directory from Step 2. Check if there's also a local vendored copy that needs updating:
 
diff --git a/gstack-upgrade/SKILL.md.tmpl b/gstack-upgrade/SKILL.md.tmpl
index ac25894b34..b6e0137f9a 100644
--- a/gstack-upgrade/SKILL.md.tmpl
+++ b/gstack-upgrade/SKILL.md.tmpl
@@ -134,7 +134,31 @@ cd "$INSTALL_DIR" && ./setup
 rm -rf "$INSTALL_DIR.bak" "$TMP_DIR"
 ```
 
-### Step 4.5: Sync local vendored copy
+### Step 4.5: Apply user overlay
+
+After `./setup` completes, check if the user has customizations that should be restored.
+
+```bash
+_OVERLAY_DIR="$HOME/.gstack/user-overlay"
+if [ -d "$_OVERLAY_DIR" ]; then
+  _SKILL_NAME=$(basename "$INSTALL_DIR")
+  if [ -d "$_OVERLAY_DIR/$_SKILL_NAME" ]; then
+    cp -Rf "$_OVERLAY_DIR/$_SKILL_NAME/." "$INSTALL_DIR/"
+    echo "USER_OVERLAY: Applied overlay for $_SKILL_NAME"
+  fi
+fi
+```
+
+If overlay files exist, they are copied over the freshly-upgraded skill directory. This preserves user customizations (edited SKILL.md files, added templates, extra scripts) across gstack updates.
+
+Users create overlays by copying files to `~/.gstack/user-overlay/<skill-name>/`. For example, to persist edits to `plan-eng-review`:
+
+```bash
+mkdir -p ~/.gstack/user-overlay/plan-eng-review
+cp ~/.claude/skills/gstack/plan-eng-review/SKILL.md.tmpl ~/.gstack/user-overlay/plan-eng-review/
+```
+
+### Step 4.6: Sync local vendored copy
 
 Use the install directory from Step 2. Check if there's also a local vendored copy that needs updating:
 
diff --git a/plan-eng-review/SKILL.md b/plan-eng-review/SKILL.md
index d2715aac09..d9a9c945b5 100644
--- a/plan-eng-review/SKILL.md
+++ b/plan-eng-review/SKILL.md
@@ -1108,9 +1108,44 @@ Check the git log for this branch. If there are prior commits suggesting a previ
 * One sentence max per option. Pick in under 5 seconds.
 * After each review section, pause and ask for feedback before moving on.
 
+## Review Artifact (persistent findings document)
+
+**PLAN MODE EXCEPTION — ALWAYS RUN:** This writes to `~/.gstack/` (user config directory), not project source files.
+
+Write the full review findings to a persistent markdown file so future sessions can recover context without re-running the review.
+
+```bash
+eval "$(~/.claude/skills/gstack/bin/gstack-slug 2>/dev/null)" 2>/dev/null || true
+mkdir -p ~/.gstack/projects/${SLUG:-unknown}/eng-reviews
+```
+
+Write to `~/.gstack/projects/{slug}/eng-reviews/{date}-{branch}-eng-review.md` using the Write tool.
+
+The file MUST contain:
+
+1. **YAML frontmatter** with `status`, `date`, `branch`, `mode`, `commit` fields
+2. **Executive Summary** — one paragraph verdict
+3. **All issues** from every review section, formatted as:
+   ```
+   ### Issue #{N}: {title}
+   - **Severity:** P0/P1/P2/P3
+   - **Confidence:** N/10
+   - **File:** `path/to/file:line`
+   - **Problem:** 2-3 sentence description
+   - **Fix:** 1-2 sentence fix
+   ```
+4. **User decisions** — which option the user chose for each AskUserQuestion
+5. **Test coverage diagram** (from Section 3)
+6. **Architecture diagram** (if any ASCII diagrams were drawn)
+7. **NOT in scope** section
+8. **Implementation order** — prioritized task list
+9. **RAM/performance notes** (if applicable)
+
+This file is the **source of truth** for future sessions. The JSONL review log is metadata for the dashboard; this file is the content for humans and AI.
+
 ## Review Log
 
-After producing the Completion Summary above, persist the review result.
+After producing the Completion Summary and Review Artifact above, persist the review result.
 
 **PLAN MODE EXCEPTION — ALWAYS RUN:** This command writes review metadata to
 `~/.gstack/` (user config directory, not project files). The skill preamble
diff --git a/plan-eng-review/SKILL.md.tmpl b/plan-eng-review/SKILL.md.tmpl
index f15fc7f589..be20e55e11 100644
--- a/plan-eng-review/SKILL.md.tmpl
+++ b/plan-eng-review/SKILL.md.tmpl
@@ -260,9 +260,44 @@ Check the git log for this branch. If there are prior commits suggesting a previ
 * One sentence max per option. Pick in under 5 seconds.
 * After each review section, pause and ask for feedback before moving on.
 
+## Review Artifact (persistent findings document)
+
+**PLAN MODE EXCEPTION — ALWAYS RUN:** This writes to `~/.gstack/` (user config directory), not project source files.
+
+Write the full review findings to a persistent markdown file so future sessions can recover context without re-running the review.
+
+```bash
+eval "$(~/.claude/skills/gstack/bin/gstack-slug 2>/dev/null)" 2>/dev/null || true
+mkdir -p ~/.gstack/projects/${SLUG:-unknown}/eng-reviews
+```
+
+Write to `~/.gstack/projects/{slug}/eng-reviews/{date}-{branch}-eng-review.md` using the Write tool.
+
+The file MUST contain:
+
+1. **YAML frontmatter** with `status`, `date`, `branch`, `mode`, `commit` fields
+2. **Executive Summary** — one paragraph verdict
+3. **All issues** from every review section, formatted as:
+   ```
+   ### Issue #{N}: {title}
+   - **Severity:** P0/P1/P2/P3
+   - **Confidence:** N/10
+   - **File:** `path/to/file:line`
+   - **Problem:** 2-3 sentence description
+   - **Fix:** 1-2 sentence fix
+   ```
+4. **User decisions** — which option the user chose for each AskUserQuestion
+5. **Test coverage diagram** (from Section 3)
+6. **Architecture diagram** (if any ASCII diagrams were drawn)
+7. **NOT in scope** section
+8. **Implementation order** — prioritized task list
+9. **RAM/performance notes** (if applicable)
+
+This file is the **source of truth** for future sessions. The JSONL review log is metadata for the dashboard; this file is the content for humans and AI.
+
 ## Review Log
 
-After producing the Completion Summary above, persist the review result.
+After producing the Completion Summary and Review Artifact above, persist the review result.
 
 **PLAN MODE EXCEPTION — ALWAYS RUN:** This command writes review metadata to
 `~/.gstack/` (user config directory, not project files). The skill preamble