diff --git a/bin/gstack-open b/bin/gstack-open
new file mode 100644
index 000000000..cae5ffea9
--- /dev/null
+++ b/bin/gstack-open
@@ -0,0 +1,37 @@
+#!/usr/bin/env bash
+# gstack-open — cross-platform URL/file opener
+# Usage: gstack-open <url-or-path>
+#
+# Replaces bare `open` calls in skill templates, which is macOS-only.
+# Works on macOS, Linux (xdg-open / sensible-browser), and Windows (Git Bash / WSL).
+set -euo pipefail
+
+TARGET="${1:-}"
+if [ -z "$TARGET" ]; then
+  echo "gstack-open: missing argument" >&2
+  exit 1
+fi
+
+case "$(uname -s)" in
+  Darwin)
+    open "$TARGET"
+    ;;
+  Linux)
+    # WSL: delegate to Windows explorer
+    if grep -qi microsoft /proc/version 2>/dev/null; then
+      explorer.exe "$TARGET" 2>/dev/null || cmd.exe /c start "" "$TARGET" 2>/dev/null || true
+    else
+      xdg-open "$TARGET" 2>/dev/null \
+        || sensible-browser "$TARGET" 2>/dev/null \
+        || x-www-browser "$TARGET" 2>/dev/null \
+        || { echo "gstack-open: no browser launcher found (tried xdg-open, sensible-browser, x-www-browser)" >&2; exit 1; }
+    fi
+    ;;
+  MINGW*|MSYS*|CYGWIN*)
+    start "" "$TARGET"
+    ;;
+  *)
+    # Unknown OS — try xdg-open as a best-effort fallback
+    xdg-open "$TARGET" 2>/dev/null || { echo "gstack-open: unsupported OS: $(uname -s)" >&2; exit 1; }
+    ;;
+esac
diff --git a/bin/gstack-review-log b/bin/gstack-review-log
index d7235bc3a..0eb9797c0 100755
--- a/bin/gstack-review-log
+++ b/bin/gstack-review-log
@@ -3,7 +3,35 @@
 # Usage: gstack-review-log '{"skill":"...","timestamp":"...","status":"..."}'
 set -euo pipefail
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+
+# Security: validate the input is a single-line JSON object before writing.
+# Reject inputs containing newlines (JSONL injection) or that are not valid JSON.
+INPUT="${1:-}"
+if [ -z "$INPUT" ]; then
+  echo "gstack-review-log: missing argument" >&2
+  exit 1
+fi
+
+# Reject multi-line input — a single JSONL entry must be one line
+if printf '%s' "$INPUT" | grep -q $'\n'; then
+  echo "gstack-review-log: input contains newlines — rejected to prevent JSONL injection" >&2
+  exit 1
+fi
+
+# Validate JSON structure using node or python if available; silently skip on failure
+if command -v node >/dev/null 2>&1; then
+  if ! node -e "JSON.parse(process.argv[1])" "$INPUT" 2>/dev/null; then
+    echo "gstack-review-log: input is not valid JSON — rejected" >&2
+    exit 1
+  fi
+elif command -v python3 >/dev/null 2>&1; then
+  if ! python3 -c "import sys,json; json.loads(sys.argv[1])" "$INPUT" 2>/dev/null; then
+    echo "gstack-review-log: input is not valid JSON — rejected" >&2
+    exit 1
+  fi
+fi
+
 eval "$("$SCRIPT_DIR/gstack-slug" 2>/dev/null)"
 GSTACK_HOME="${GSTACK_HOME:-$HOME/.gstack}"
 mkdir -p "$GSTACK_HOME/projects/$SLUG"
-echo "$1" >> "$GSTACK_HOME/projects/$SLUG/$BRANCH-reviews.jsonl"
+printf '%s\n' "$INPUT" >> "$GSTACK_HOME/projects/$SLUG/$BRANCH-reviews.jsonl"
diff --git a/bin/gstack-telemetry-log b/bin/gstack-telemetry-log
index 5cddc519f..84cdd2d0a 100755
--- a/bin/gstack-telemetry-log
+++ b/bin/gstack-telemetry-log
@@ -151,15 +151,37 @@ fi
 # ─── Construct and append JSON ───────────────────────────────
 mkdir -p "$ANALYTICS_DIR"
 
+# Sanitize a string for safe embedding in a JSONL value:
+# - strip newlines/carriage returns (prevent JSONL injection — extra lines in the file)
+# - escape backslashes and double-quotes (prevent JSON string corruption)
+# - truncate to $2 bytes (optional, default 200)
+sanitize_str() {
+  local val="$1"
+  local maxlen="${2:-200}"
+  printf '%s' "$val" \
+    | tr -d '\n\r' \
+    | head -c "$maxlen" \
+    | sed 's/\\/\\\\/g; s/"/\\"/g'
+}
+
 # Escape null fields
 ERR_FIELD="null"
-[ -n "$ERROR_CLASS" ] && ERR_FIELD="\"$ERROR_CLASS\""
+[ -n "$ERROR_CLASS" ] && ERR_FIELD="\"$(sanitize_str "$ERROR_CLASS" 100)\""
 
 ERR_MSG_FIELD="null"
-[ -n "$ERROR_MESSAGE" ] && ERR_MSG_FIELD="\"$(echo "$ERROR_MESSAGE" | head -c 200 | sed 's/"/\\"/g')\""
+[ -n "$ERROR_MESSAGE" ] && ERR_MSG_FIELD="\"$(sanitize_str "$ERROR_MESSAGE" 200)\""
 
 STEP_FIELD="null"
-[ -n "$FAILED_STEP" ] && STEP_FIELD="\"$(echo "$FAILED_STEP" | head -c 30)\""
+[ -n "$FAILED_STEP" ] && STEP_FIELD="\"$(sanitize_str "$FAILED_STEP" 30)\""
+
+# Sanitize fields that appear directly in the printf format string
+SKILL="$(sanitize_str "$SKILL" 60)"
+SESSION_ID="$(sanitize_str "$SESSION_ID" 60)"
+EVENT_TYPE="$(sanitize_str "$EVENT_TYPE" 40)"
+OUTCOME="$(sanitize_str "$OUTCOME" 20)"
+SOURCE="$(sanitize_str "$SOURCE" 20)"
+REPO_SLUG="$(sanitize_str "$REPO_SLUG" 120)"
+BRANCH="$(sanitize_str "$BRANCH" 120)"
 
 # Cap unreasonable durations
 if [ -n "$DURATION" ] && [ "$DURATION" -gt 86400 ] 2>/dev/null; then
diff --git a/codex/SKILL.md.tmpl b/codex/SKILL.md.tmpl
index c0b7adb1b..627cf8382 100644
--- a/codex/SKILL.md.tmpl
+++ b/codex/SKILL.md.tmpl
@@ -80,13 +80,13 @@ TMPERR=$(mktemp /tmp/codex-err-XXXXXX.txt)
 
 2. Run the review (5-minute timeout):
 ```bash
-codex review --base <base> -c 'model_reasoning_effort="xhigh"' --enable web_search_cached 2>"$TMPERR"
+codex review --base <base> -c 'model_reasoning_effort="high"' --search 2>"$TMPERR"
 ```
 
 Use `timeout: 300000` on the Bash call. If the user provided custom instructions
 (e.g., `/codex review focus on security`), pass them as the prompt argument:
 ```bash
-codex review "focus on security" --base <base> -c 'model_reasoning_effort="xhigh"' --enable web_search_cached 2>"$TMPERR"
+codex review "focus on security" --base <base> -c 'model_reasoning_effort="high"' --search 2>"$TMPERR"
 ```
 
 3. Capture the output. Then parse cost from stderr:
@@ -95,8 +95,8 @@ grep "tokens used" "$TMPERR" 2>/dev/null || echo "tokens: unknown"
 ```
 
 4. Determine gate verdict by checking the review output for critical findings.
-   If the output contains `[P1]` — the gate is **FAIL**.
-   If no `[P1]` markers are found (only `[P2]` or no findings) — the gate is **PASS**.
+   If the output contains `[P0]` or `[P1]` — the gate is **FAIL**.
+   If no `[P0]` or `[P1]` markers are found (only `[P2]` or no findings) — the gate is **PASS**.
 
 5. Present the output:
 
@@ -131,7 +131,7 @@ CROSS-MODEL ANALYSIS:
 ```
 
 Substitute: TIMESTAMP (ISO 8601), STATUS ("clean" if PASS, "issues_found" if FAIL),
-GATE ("pass" or "fail"), findings (count of [P1] + [P2] markers),
+GATE ("pass" or "fail"), findings (count of [P0] + [P1] + [P2] markers),
 findings_fixed (count of findings that were addressed/fixed before shipping).
 
 8. Clean up temp files:
@@ -159,7 +159,7 @@ With focus (e.g., "security"):
 
 2. Run codex exec with **JSONL output** to capture reasoning traces and tool calls (5-minute timeout):
 ```bash
-codex exec "<prompt>" -s read-only -c 'model_reasoning_effort="xhigh"' --enable web_search_cached --json 2>/dev/null | python3 -c "
+codex exec "<prompt>" -s read-only -c 'model_reasoning_effort="high"' --search --json 2>/dev/null | python3 -c "
 import sys, json
 for line in sys.stdin:
     line = line.strip()
@@ -244,7 +244,7 @@ THE PLAN:
 
 For a **new session:**
 ```bash
-codex exec "<prompt>" -s read-only -c 'model_reasoning_effort="xhigh"' --enable web_search_cached --json 2>"$TMPERR" | python3 -c "
+codex exec "<prompt>" -s read-only -c 'model_reasoning_effort="high"' --search --json 2>"$TMPERR" | python3 -c "
 import sys, json
 for line in sys.stdin:
     line = line.strip()
@@ -277,7 +277,7 @@ for line in sys.stdin:
 
 For a **resumed session** (user chose "Continue"):
 ```bash
-codex exec resume <session-id> "<prompt>" -s read-only -c 'model_reasoning_effort="xhigh"' --enable web_search_cached --json 2>"$TMPERR" | python3 -c "
+codex exec resume <session-id> "<prompt>" -s read-only -c 'model_reasoning_effort="high"' --search --json 2>"$TMPERR" | python3 -c "
 <same python streaming parser as above>
 "
 ```
@@ -313,10 +313,10 @@ Session saved — run /codex again to continue this conversation.
 agentic coding model). This means as OpenAI ships newer models, /codex automatically
 uses them. If the user wants a specific model, pass `-m` through to codex.
 
-**Reasoning effort:** All modes use `xhigh` — maximum reasoning power. When reviewing code, breaking code, or consulting on architecture, you want the model thinking as hard as possible.
+**Reasoning effort:** All modes use `high` — maximum reasoning power. When reviewing code, breaking code, or consulting on architecture, you want the model thinking as hard as possible. (`xhigh` is not a valid value and causes an API error.)
 
-**Web search:** All codex commands use `--enable web_search_cached` so Codex can look up
-docs and APIs during review. This is OpenAI's cached index — fast, no extra cost.
+**Web search:** All codex commands use `--search` so Codex can look up
+docs and APIs during review. (`--enable web_search_cached` is deprecated — use `--search`.)
 
 If the user specifies a model (e.g., `/codex review -m gpt-5.1-codex-max`
 or `/codex challenge -m gpt-5.2`), pass the `-m` flag through to codex.
diff --git a/office-hours/SKILL.md.tmpl b/office-hours/SKILL.md.tmpl
index 93abb1bb6..8f432c6df 100644
--- a/office-hours/SKILL.md.tmpl
+++ b/office-hours/SKILL.md.tmpl
@@ -600,7 +600,7 @@ Say:
 
 Then use AskUserQuestion: "Would you consider applying to Y Combinator?"
 
-- If yes → run `open https://ycombinator.com/apply?ref=gstack` and say: "Bring this design doc to your YC interview. It's better than most pitch decks."
+- If yes → run `~/.claude/skills/gstack/bin/gstack-open https://ycombinator.com/apply?ref=gstack` and say: "Bring this design doc to your YC interview. It's better than most pitch decks."
 - If no → respond warmly: "Totally fair. The design doc is yours either way — and the offer stands if you ever change your mind." Then proceed to next-skill recs. No pressure, no guilt, no re-ask.
 
 **Middle tier** — emotional target: *"I might be onto something."* Validation + curiosity.
diff --git a/scripts/gen-skill-docs.ts b/scripts/gen-skill-docs.ts
index 8d483dad2..bc0fcc696 100644
--- a/scripts/gen-skill-docs.ts
+++ b/scripts/gen-skill-docs.ts
@@ -235,7 +235,7 @@ thing when AI makes the marginal cost near-zero. Read more: https://garryslist.o
 Then offer to open the essay in their default browser:
 
 \`\`\`bash
-open https://garryslist.org/posts/boil-the-ocean
+~/.claude/skills/gstack/bin/gstack-open https://garryslist.org/posts/boil-the-ocean
 touch ~/.gstack/.completeness-intro-seen
 \`\`\`
 
diff --git a/scripts/resolvers/preamble.ts b/scripts/resolvers/preamble.ts
index 76573422e..4d9789e84 100644
--- a/scripts/resolvers/preamble.ts
+++ b/scripts/resolvers/preamble.ts
@@ -61,7 +61,7 @@ thing when AI makes the marginal cost near-zero. Read more: https://garryslist.o
 Then offer to open the essay in their default browser:
 
 \`\`\`bash
-open https://garryslist.org/posts/boil-the-ocean
+~/.claude/skills/gstack/bin/gstack-open https://garryslist.org/posts/boil-the-ocean
 touch ~/.gstack/.completeness-intro-seen
 \`\`\`
 
diff --git a/setup b/setup
index bfae87851..38893a438 100755
--- a/setup
+++ b/setup
@@ -3,9 +3,20 @@
 set -e
 
 if ! command -v bun >/dev/null 2>&1; then
-  echo "Error: bun is required but not installed." >&2
-  echo "Install it: curl -fsSL https://bun.sh/install | bash" >&2
-  exit 1
+  echo "Bun not found — installing automatically..."
+  if ! curl -fsSL https://bun.sh/install | bash; then
+    echo "Error: Bun auto-install failed." >&2
+    echo "Install it manually: curl -fsSL https://bun.sh/install | bash" >&2
+    exit 1
+  fi
+  # Pick up the newly installed bun from the default location
+  export PATH="$HOME/.bun/bin:$PATH"
+  if ! command -v bun >/dev/null 2>&1; then
+    echo "Error: bun installed but not found in PATH." >&2
+    echo "Open a new terminal and re-run setup, or add ~/.bun/bin to your PATH." >&2
+    exit 1
+  fi
+  echo "Bun installed successfully."
 fi
 
 INSTALL_GSTACK_DIR="$(cd "$(dirname "$0")" && pwd)"