diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
index 73faddd..a23d9eb 100644
--- a/.claude-plugin/plugin.json
+++ b/.claude-plugin/plugin.json
@@ -1,6 +1,6 @@
 {
   "name": "jsc-log",
-  "version": "0.0.8",
+  "version": "0.0.9",
   "description": "工作日誌（LOG_{HASH} wiki 頁）、技能使用統計與教訓紀錄（LEARN_{HASH} wiki 頁）",
   "skills": "./skills",
   "author": {
diff --git a/.codex-plugin/plugin.json b/.codex-plugin/plugin.json
index 09756e6..4e99326 100644
--- a/.codex-plugin/plugin.json
+++ b/.codex-plugin/plugin.json
@@ -1,6 +1,6 @@
 {
   "name": "jsc-log",
-  "version": "0.0.8",
+  "version": "0.0.9",
   "description": "工作日誌（LOG_{HASH} wiki 頁）、技能使用統計與教訓紀錄（LEARN_{HASH} wiki 頁）",
   "skills": "./skills"
 }
diff --git a/README.md b/README.md
index 3335079..86439b0 100644
--- a/README.md
+++ b/README.md
@@ -24,6 +24,7 @@ Marketplace 統一為 `jsc`（https://gitea.jsc.idv.tw/plugins/meta.git），安
 | --- | --- |
 | `tools/usage-stats.sh` | 聚合 `$JSC_HOME/usage/*.jsonl`：`skills` 列技能使用次數、`chains` 列呼叫鏈次數（皆降冪），`--cli <name>` 過濾 |
 | `tools/worklog-target.sh` | 接收已由 `jsc-gitea/tools/hash-id` 算好的 `HASH`，組出 `LOG_{HASH}`、`LOG_CONTENTS`（本身不再計算 SHA-1） |
+| `tools/worklog-pending.sh` | 待寫入日誌的暫存區，存放於 `$JSC_HOME/worklog-pending/{HASH}/`。`add {HASH} {檔案}` 存一段內容（`jsc-sdlc` 的階段回報發現沒寫日誌時會呼叫），`cat {HASH}` 依時間印出全部、`list` 列路徑、`clear` 清除。結束碼 `3` 代表沒有暫存內容。**寫進 wiki 成功之後才 clear**，先清再寫會兩邊都沒有 |
 | `tools/token-usage.sh` | 讀單一 CLI 這次工作的 token 用量，印出「input(tab)output」；來源讀不到就印「N/A(tab)N/A」並正常結束。第二個參數傳 session id，就只讀該階段的 transcript，數字才會跟花費時間對得上。各 CLI 的取得方式寫在腳本開頭註解 |
 
 ## Skills 目錄
@@ -34,7 +35,7 @@ Marketplace 統一為 `jsc`（https://gitea.jsc.idv.tw/plugins/meta.git），安
 
 ### `worklog`
 
-工作完成後蒐集十項資訊（存取庫、分支、計畫連結、工作包連結、花費時間、token 用量、任務狀態、執行細節、困難與解決、PR 目標分支），用 `tools/worklog-target.sh` 產生目標頁，再套範本附加到 `LOG_{HASH}` 與 `LOG_CONTENTS`。
+工作完成後蒐集十項資訊（存取庫、分支、計畫連結、工作包連結、花費時間、token 用量、任務狀態、執行細節、困難與解決、PR 目標分支），用 `tools/worklog-target.sh` 產生目標頁，再套範本附加到 `LOG_{HASH}` 與 `LOG_CONTENTS`。寫入前先讀 `tools/worklog-pending.sh cat {HASH}`：之前有階段跑完沒寫日誌，內容會暫存在那裡，這次一併寫進去，**寫成功之後才清暫存**。
 
 ### `stats`
 
diff --git a/plugin.json b/plugin.json
index bbed3be..1c557aa 100644
--- a/plugin.json
+++ b/plugin.json
@@ -1,6 +1,6 @@
 {
   "name": "jsc-log",
-  "version": "0.0.8",
+  "version": "0.0.9",
   "description": "工作日誌（LOG_{HASH} wiki 頁）、技能使用統計與教訓紀錄（LEARN_{HASH} wiki 頁）",
   "skills": "./skills/"
 }
diff --git a/skills/worklog/SKILL.md b/skills/worklog/SKILL.md
index e7aa0e3..9c2b4fe 100644
--- a/skills/worklog/SKILL.md
+++ b/skills/worklog/SKILL.md
@@ -1,6 +1,6 @@
 ---
 name: worklog
-description: After finishing a work package, collect ten facts (repo, branch, plan link, work package link, elapsed time from session-timer, token usage per CLI, status, details, difficulties, PR target) and append a templated entry to wiki LOG_{HASH} plus LOG_CONTENTS. HASH follows the shared 8-char rule with the H-prefix fallback, and the work-week Friday still drives the page content. Trigger at the end of implement or maintain; not for planning notes.
+description: After finishing a work package, collect ten facts (repo, branch, plan link, work package link, elapsed time from session-timer, token usage per CLI, status, details, difficulties, PR target) and append a templated entry to wiki LOG_{HASH} plus LOG_CONTENTS. HASH follows the shared 8-char rule with the H-prefix fallback, and the work-week Friday still drives the page content. Merge whatever tools/worklog-pending.sh holds for that HASH into the same write, then clear the pending area once the wiki write succeeded. Trigger at the end of implement or maintain; not for planning notes.
 ---
 
 # worklog — work log
@@ -30,5 +30,7 @@ The `{HASH}` in every page name above is computed with `jsc-gitea/tools/hash-id`
 2. Compute `{HASH}` from the code repo's `{owner}/{repo}` with `jsc-gitea/tools/hash-id`. Done when the 8-character `{HASH}` is known.
 3. Run `tools/worklog-target.sh "{HASH}" all`. Use `PAGE` for `LOG_{HASH}` and `CONTENTS` for `LOG_CONTENTS`. Done when both page names are known.
 4. Fix the work week: the Friday of the current work week drives the page content and the row dates. Done when that Friday is fixed as a `yyyy-MM-dd` date.
-5. Read `PAGE` via `jsc-gitea:wiki`. If it does not exist, create it with the structure described in `templates/log-entry.md`; otherwise APPEND the new entry at the end and never overwrite existing entries. Done when the new entry exists on `PAGE`.
-6. Update `CONTENTS` in the same pass (apply `templates/log-contents.md`; add the row if missing, otherwise refresh its 條目數 and 最後更新). Done when the row for `PAGE` carries this week's Friday date.
+5. Collect what the pending area holds for this `{HASH}`: run `tools/worklog-pending.sh cat {HASH}`. Exit 3 means nothing is pending — carry on with this run's entry alone. Anything it prints was written by an earlier SDLC stage that ended without a work log, so it goes into **this** write, ahead of this run's own entry, in the order printed. Done when the pending content is either merged into the entries about to be written, or confirmed empty.
+6. Read `PAGE` via `jsc-gitea:wiki`. If it does not exist, create it with the structure described in `templates/log-entry.md`; otherwise APPEND the new entries at the end and never overwrite existing entries. Done when every entry from step 5 plus this run's own entry exists on `PAGE`.
+7. Update `CONTENTS` in the same pass (apply `templates/log-contents.md`; add the row if missing, otherwise refresh its 條目數 and 最後更新). Done when the row for `PAGE` carries this week's Friday date.
+8. Clear the pending area: run `tools/worklog-pending.sh clear {HASH}` **only after the wiki write of step 6 succeeded**. Clearing first and failing the write loses the content on both sides. Skip this when step 5 found nothing. Done when the script reports the cleared path, or step 5 was empty.
diff --git a/tools/worklog-pending.sh b/tools/worklog-pending.sh
new file mode 100755
index 0000000..a530b72
--- /dev/null
+++ b/tools/worklog-pending.sh
@@ -0,0 +1,92 @@
+#!/usr/bin/env sh
+# worklog-pending.sh — 待寫入工作日誌的暫存區（供 jsc-sdlc 階段回報與 jsc-log:worklog 使用）。
+#
+# 為什麼要有這支腳本：SDLC 每個階段結束都要回報工作日誌連結。階段跑完卻沒寫日誌時，
+# 內容如果只留在對話裡，換一個工作階段就消失了，下次寫日誌也補不回來。所以暫存搬到檔案層：
+# 階段回報發現沒有日誌，就把該階段的內容存進本區並警告使用者；下次 jsc-log:worklog 寫入時
+# 先把本區的內容一併寫進去，寫完才清掉。
+#
+# 用法:
+#   worklog-pending.sh add <hash> <file>   把一段待寫入的日誌內容存起來，印出存放路徑
+#   worklog-pending.sh list <hash>         列出該 hash 的暫存檔路徑（一行一個，依時間排序）
+#   worklog-pending.sh cat <hash>          依時間順序印出全部暫存內容
+#   worklog-pending.sh clear <hash>        清掉該 hash 的暫存（寫入日誌成功後才做）
+#
+# <hash> 為 jsc-gitea/tools/hash-id 算出的 8 碼工作日誌 hash，也就是 LOG_{HASH} 的 HASH。
+# 存放位置: $JSC_HOME/worklog-pending/{hash}/{UTC 時間}-{pid}.md（JSC_HOME 預設 ~/.jsc）
+#
+# 結束碼: 0=成功 1=讀寫失敗 2=用法錯誤 3=該 hash 沒有暫存內容
+#
+# 陷阱:
+#   - clear 只在日誌確實寫進 wiki 之後才呼叫。先清再寫，寫失敗就兩邊都沒有。
+#   - 檔名帶 UTC 時間與 pid，同一秒內兩個工作階段各存各的，不會互相覆蓋。
+set -eu
+
+JSC_HOME="${JSC_HOME:-$HOME/.jsc}"
+ROOT="$JSC_HOME/worklog-pending"
+
+usage() {
+  cat >&2 <<'EOF'
+用法:
+  worklog-pending.sh add <hash> <file>   存一段待寫入的日誌內容
+  worklog-pending.sh list <hash>         列出暫存檔路徑
+  worklog-pending.sh cat <hash>          印出全部暫存內容
+  worklog-pending.sh clear <hash>        清掉暫存（寫入日誌成功後才做）
+結束碼: 0=成功 1=讀寫失敗 2=用法錯誤 3=沒有暫存內容
+EOF
+  exit 2
+}
+
+valid_hash() { # 只收 8 碼大寫英數，擋掉路徑穿越
+  case "$1" in
+    [0-9A-Z][0-9A-Z][0-9A-Z][0-9A-Z][0-9A-Z][0-9A-Z][0-9A-Z][0-9A-Z]) return 0 ;;
+    *) echo "[jsc][工作日誌暫存][ERR]：hash 須為 8 碼大寫英數，收到「${1:-空值}」。" >&2; exit 2 ;;
+  esac
+}
+
+cmd="${1:-}"; [ -n "$cmd" ] || usage
+shift || true
+hash="${1:-}"; [ -n "$hash" ] || usage
+valid_hash "$hash"
+dir="$ROOT/$hash"
+
+case "$cmd" in
+  add)
+    src="${2:-}"
+    [ -n "$src" ] || usage
+    [ -f "$src" ] || { echo "[jsc][工作日誌暫存][ERR]：找不到內容檔「$src」。" >&2; exit 1; }
+    mkdir -p "$dir" || { echo "[jsc][工作日誌暫存][ERR]：建不出暫存目錄「$dir」。" >&2; exit 1; }
+    stamp=$(date -u +%Y%m%dT%H%M%SZ)
+    target="$dir/$stamp-$$.md"
+    cp "$src" "$target" || { echo "[jsc][工作日誌暫存][ERR]：寫不進「$target」。" >&2; exit 1; }
+    printf '%s\n' "$target"
+    ;;
+  list)
+    [ -d "$dir" ] || exit 3
+    found=0
+    for f in "$dir"/*.md; do
+      [ -f "$f" ] || continue
+      printf '%s\n' "$f"
+      found=1
+    done
+    [ "$found" -eq 1 ] || exit 3
+    ;;
+  cat)
+    [ -d "$dir" ] || exit 3
+    found=0
+    for f in "$dir"/*.md; do
+      [ -f "$f" ] || continue
+      cat "$f"
+      printf '\n'
+      found=1
+    done
+    [ "$found" -eq 1 ] || exit 3
+    ;;
+  clear)
+    [ -d "$dir" ] || exit 3
+    rm -rf "$dir" || { echo "[jsc][工作日誌暫存][ERR]：清不掉「$dir」。" >&2; exit 1; }
+    printf '已清除暫存：%s\n' "$dir"
+    ;;
+  *)
+    usage ;;
+esac