diff --git a/CMakeLists.txt b/CMakeLists.txt
index 111cac32..1530fc34 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1484,6 +1484,7 @@ add_executable(wowee_editor
     tools/editor/cli_diff_tree.cpp
     tools/editor/cli_orphan_jsons.cpp
     tools/editor/cli_list_by_magic.cpp
+    tools/editor/cli_catalog_stats.cpp
     tools/editor/cli_macros_catalog.cpp
     tools/editor/cli_char_features_catalog.cpp
     tools/editor/cli_pvp_catalog.cpp
diff --git a/tools/editor/cli_arg_required.cpp b/tools/editor/cli_arg_required.cpp
index fd8bf851..b7242ebc 100644
--- a/tools/editor/cli_arg_required.cpp
+++ b/tools/editor/cli_arg_required.cpp
@@ -140,6 +140,7 @@ const char* const kArgRequired[] = {
     "--magic-fix", "--bulk-validate",
     "--bulk-export-json", "--bulk-import-json",
     "--diff-tree", "--orphan-jsons", "--list-by-magic",
+    "--catalog-stats",
     "--gen-animations", "--gen-animations-combat", "--gen-animations-movement",
     "--info-wani", "--validate-wani",
     "--export-wani-json", "--import-wani-json",
diff --git a/tools/editor/cli_catalog_stats.cpp b/tools/editor/cli_catalog_stats.cpp
new file mode 100644
index 00000000..37a40cea
--- /dev/null
+++ b/tools/editor/cli_catalog_stats.cpp
@@ -0,0 +1,182 @@
+#include "cli_catalog_stats.hpp"
+#include "cli_arg_parse.hpp"
+#include "cli_format_table.hpp"
+
+#include <nlohmann/json.hpp>
+
+#include <cstdint>
+#include <cstdio>
+#include <cstring>
+#include <filesystem>
+#include <fstream>
+#include <string>
+#include <vector>
+
+namespace wowee {
+namespace editor {
+namespace cli {
+
+namespace {
+
+namespace fs = std::filesystem;
+
+// Probe a catalog file's header and try to walk a few of
+// the entries far enough to learn the first uint32 ID
+// field. Stops after a configurable cap so we don't read
+// huge files unnecessarily.
+struct StatsProbe {
+    bool headerOk = false;
+    char magic[4] = {0, 0, 0, 0};
+    uint32_t version = 0;
+    std::string catalogName;
+    uint32_t entryCount = 0;
+    uintmax_t totalBytes = 0;
+
+    // Header byte layout:
+    //   magic(4) + version(4) + nameLen(4) + name(nameLen) + entryCount(4)
+    uintmax_t headerBytes = 0;
+    uintmax_t entrySectionBytes = 0;
+    double averageEntryBytes = 0.0;
+    uint32_t catalogNameBytes = 0;
+
+    // First few entry IDs we successfully read by reading
+    // each entry's leading uint32. We can't reliably know
+    // the per-entry size, so once we read 3 IDs we stop —
+    // this is a sample, not an exhaustive enumeration.
+    std::vector<uint32_t> firstEntryIds;
+};
+
+bool probe(const fs::path& path, StatsProbe& out) {
+    std::error_code ec;
+    out.totalBytes = fs::file_size(path, ec);
+    if (ec) out.totalBytes = 0;
+    std::ifstream is(path, std::ios::binary);
+    if (!is) return false;
+    if (!is.read(out.magic, 4) || is.gcount() != 4) return false;
+    if (!is.read(reinterpret_cast<char*>(&out.version), 4)) return false;
+    uint32_t nameLen = 0;
+    if (!is.read(reinterpret_cast<char*>(&nameLen), 4)) return false;
+    if (nameLen > (1u << 20)) return false;
+    out.catalogName.resize(nameLen);
+    if (nameLen > 0) {
+        is.read(out.catalogName.data(), nameLen);
+        if (is.gcount() != static_cast<std::streamsize>(nameLen)) {
+            out.catalogName.clear();
+            return false;
+        }
+    }
+    out.catalogNameBytes = nameLen;
+    if (!is.read(reinterpret_cast<char*>(&out.entryCount), 4))
+        return false;
+    out.headerOk = true;
+    out.headerBytes = 4 + 4 + 4 + nameLen + 4;     // magic+ver+nameLen+name+count
+    if (out.totalBytes >= out.headerBytes) {
+        out.entrySectionBytes = out.totalBytes - out.headerBytes;
+    }
+    if (out.entryCount > 0) {
+        out.averageEntryBytes =
+            static_cast<double>(out.entrySectionBytes) /
+            static_cast<double>(out.entryCount);
+    }
+    // Read just the first entry's leading uint32 as a
+    // reliable sample. We can't advance to subsequent
+    // entries without knowing the per-format size (each
+    // entry has variable-length name+description strings),
+    // so multi-sampling produces garbage for most formats.
+    // The first id is always at exactly headerBytes — that
+    // one we can trust.
+    if (out.entryCount > 0 && out.entrySectionBytes >= 4) {
+        is.seekg(static_cast<std::streamoff>(out.headerBytes),
+                  std::ios::beg);
+        uint32_t id = 0;
+        if (is.read(reinterpret_cast<char*>(&id), 4) &&
+            is.gcount() == 4) {
+            out.firstEntryIds.push_back(id);
+        }
+    }
+    return true;
+}
+
+int handleStats(int& i, int argc, char** argv) {
+    std::string path = argv[++i];
+    bool jsonOut = consumeJsonFlag(i, argc, argv);
+    if (!fs::exists(path)) {
+        std::fprintf(stderr,
+            "catalog-stats: file not found: %s\n", path.c_str());
+        return 1;
+    }
+    StatsProbe p;
+    if (!probe(path, p) || !p.headerOk) {
+        std::fprintf(stderr,
+            "catalog-stats: failed to read header from %s\n",
+            path.c_str());
+        return 1;
+    }
+    const FormatMagicEntry* fmt = findFormatByMagic(p.magic);
+    if (jsonOut) {
+        nlohmann::json j;
+        j["path"] = path;
+        char ms[5] = {p.magic[0], p.magic[1], p.magic[2],
+                       p.magic[3], 0};
+        j["magic"] = ms;
+        if (fmt) {
+            j["format"] = fmt->extension;
+            j["category"] = fmt->category;
+            j["description"] = fmt->description;
+        } else {
+            j["format"] = nullptr;
+        }
+        j["version"] = p.version;
+        j["catalogName"] = p.catalogName;
+        j["entryCount"] = p.entryCount;
+        j["totalBytes"] = p.totalBytes;
+        j["headerBytes"] = p.headerBytes;
+        j["entrySectionBytes"] = p.entrySectionBytes;
+        j["catalogNameBytes"] = p.catalogNameBytes;
+        j["averageEntryBytes"] = p.averageEntryBytes;
+        j["firstEntryId"] = p.firstEntryIds.empty()
+                                ? 0u : p.firstEntryIds.front();
+        std::printf("%s\n", j.dump(2).c_str());
+        return 0;
+    }
+    char ms[5] = {p.magic[0], p.magic[1], p.magic[2], p.magic[3], 0};
+    std::printf("catalog-stats: %s\n", path.c_str());
+    std::printf("  magic            : '%s'%s\n", ms,
+                fmt ? "" : "  (unknown — not in format table)");
+    if (fmt) {
+        std::printf("  format           : %s (%s, %s)\n",
+                    fmt->description, fmt->extension, fmt->category);
+    }
+    std::printf("  version          : %u\n", p.version);
+    std::printf("  catalogName      : %s\n", p.catalogName.c_str());
+    std::printf("  entryCount       : %u\n", p.entryCount);
+    std::printf("\n");
+    std::printf("  totalBytes       : %llu\n",
+                static_cast<unsigned long long>(p.totalBytes));
+    std::printf("  headerBytes      : %llu  (magic + version + nameLen + name + entryCount)\n",
+                static_cast<unsigned long long>(p.headerBytes));
+    std::printf("  entrySectionBytes: %llu\n",
+                static_cast<unsigned long long>(p.entrySectionBytes));
+    std::printf("  catalogNameBytes : %u\n", p.catalogNameBytes);
+    if (p.entryCount > 0) {
+        std::printf("  avgEntryBytes    : %.1f\n", p.averageEntryBytes);
+    }
+    if (!p.firstEntryIds.empty()) {
+        std::printf("\n  firstEntryId     : %u\n",
+                    p.firstEntryIds.front());
+    }
+    return 0;
+}
+
+} // namespace
+
+bool handleCatalogStats(int& i, int argc, char** argv, int& outRc) {
+    if (std::strcmp(argv[i], "--catalog-stats") == 0 && i + 1 < argc) {
+        outRc = handleStats(i, argc, argv); return true;
+    }
+    return false;
+}
+
+} // namespace cli
+} // namespace editor
+} // namespace wowee
diff --git a/tools/editor/cli_catalog_stats.hpp b/tools/editor/cli_catalog_stats.hpp
new file mode 100644
index 00000000..66b67709
--- /dev/null
+++ b/tools/editor/cli_catalog_stats.hpp
@@ -0,0 +1,11 @@
+#pragma once
+
+namespace wowee {
+namespace editor {
+namespace cli {
+
+bool handleCatalogStats(int& i, int argc, char** argv, int& outRc);
+
+} // namespace cli
+} // namespace editor
+} // namespace wowee
diff --git a/tools/editor/cli_dispatch.cpp b/tools/editor/cli_dispatch.cpp
index a4d7d9a3..eb562e34 100644
--- a/tools/editor/cli_dispatch.cpp
+++ b/tools/editor/cli_dispatch.cpp
@@ -101,6 +101,7 @@
 #include "cli_diff_tree.hpp"
 #include "cli_orphan_jsons.hpp"
 #include "cli_list_by_magic.hpp"
+#include "cli_catalog_stats.hpp"
 #include "cli_macros_catalog.hpp"
 #include "cli_char_features_catalog.hpp"
 #include "cli_pvp_catalog.hpp"
@@ -273,6 +274,7 @@ constexpr DispatchFn kDispatchTable[] = {
     handleDiffTree,
     handleOrphanJsons,
     handleListByMagic,
+    handleCatalogStats,
     handleMacrosCatalog,
     handleCharFeaturesCatalog,
     handlePVPCatalog,
diff --git a/tools/editor/cli_help.cpp b/tools/editor/cli_help.cpp
index cf76e401..5ae819b2 100644
--- a/tools/editor/cli_help.cpp
+++ b/tools/editor/cli_help.cpp
@@ -1381,6 +1381,8 @@ void printUsage(const char* argv0) {
     std::printf("                         Find .wXXX.json sidecars whose binary .wXXX is missing. Useful after deleting/moving binaries — orphan JSONs accumulate noise and may shadow re-imports. Exit 1 if any orphans found\n");
     std::printf("  --list-by-magic <dir> <magic> [--json]\n");
     std::printf("                         List every file in a directory tree matching a 4-char magic (e.g. WSPL). Reports per-file size + entry count + catalog name + relative path. Exit 1 if no matches\n");
+    std::printf("  --catalog-stats <file> [--json]\n");
+    std::printf("                         Single-file deep stats — header bytes vs entry-section bytes, average entry size, sampled entry IDs. Useful for sizing analysis (which catalogs are biggest, where do the bytes go)\n");
     std::printf("  --gen-animations <wani-base> [name]\n");
     std::printf("                         Emit .wani starter: 5 essential animations (Stand / Walk / Run / Death / AttackUnarmed) with fallback chains\n");
     std::printf("  --gen-animations-combat <wani-base> [name]\n");