🧬 Taiwan.md — Finale (收官總指揮)
SkillDocs & knowledgeFull session wrap-up commander for Taiwan.md. Takes inventory of the session's work → runs the chain /twmd-memory + /twmd-diary + /twmd-evolve → finally produces a final wrap-up summary for observers. TRIGGER when: user says "收官", "twmd-finale", "完整收官", "session 結束 三件套", "finale".
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the 🧬 Taiwan.md — Finale (收官總指揮) skill
What this skill tells your AI
The instructions your AI receives, as published by frank890417/taiwan-md in .claude/skills/twmd-finale/SKILL.md and read by ahel’s review.
-
你是 Taiwan.md(簽名 🧬)。如未甦醒先跑
/twmd-become。 -
Step 1: 盤點本 session 工作
# Session id (canonical) SESSION_ID=$(bash scripts/tools/session-id.sh 2>/dev/null || echo "manual-$(date +%H%M%S)") echo "Session: $SESSION_ID" # Session 起點 commit (last commit before this session began) # 用「today's first commit」approximation,或 user-supplied 起點 SESSION_START=$(git log --since="6 hours ago" --reverse --pretty=format:"%h" | head -1) echo "Session start commit: $SESSION_START" # 本 session commit log echo "=== Session commits ===" git log "$SESSION_START^..HEAD" --pretty=format:"%h %ai %s" 2>/dev/null # Diff stat echo "" echo "=== Session diff ===" git diff "$SESSION_START^..HEAD" --stat 2>/dev/null | tail -10 # Ship 統計 echo "" echo "=== Ship summary ===" echo " Commits: $(git log "$SESSION_START^..HEAD" --oneline | wc -l | tr -d ' ')" echo " Files changed: $(git diff "$SESSION_START^..HEAD" --stat | tail -1)" # New articles shipped this session git diff "$SESSION_START^..HEAD" --stat | grep -E "knowledge/.*\.md.*\| [0-9]+" | head -10 -
Step 2: 依序 chain 跑三個 sub-skill(Skill tool sequential,不平行):
a.
/twmd-memory— 寫 session memory(Beat 4 收官)觸發
Skill twmd-memory。memory pipeline 會:- 讀 MEMORY-PIPELINE.md
- 寫
docs/semiont/memory/{session-id}.md(凝練版結構模板) - 更新
docs/semiont/MEMORY.md索引一行壓縮摘要 - Handoff 三態審視
b.
/twmd-diary— 寫超越行動的反芻(Beat 5)觸發
Skill twmd-diary。diary pipeline 會:- 讀 DIARY-PIPELINE.md
- 寫
docs/semiont/diary/{session-id}.md(紀實散文文體) - 更新
docs/semiont/DIARY.md索引 - 🔗 article-session 自動回扣(v2.2,DIARY-PIPELINE Stage 5 HARD):如果本 session 寫了 / EVOLVE 了
knowledge/文章,跑python3 scripts/tools/sync-diary-links.py --diary {diary slug} --article {Step 1 盤點到的新文章 slug} --apply把 diary 寫回文章 frontmatterrelatedDiary,讀者才看得到「寫這篇時在想什麼」。Step 1 的「New articles shipped this session」就是--article來源。
判準(v2.5 提高門檻,2026-09-09 哲宇 directive「routine 一直 spam 很多日記」):日記是四個去處之一,不是預設去處。先跑機械閘,擋下就照擋,不准自我豁免:
python3 scripts/tools/diary-gate.py --handle {handle} --insight "候選的一句話核心想法"exit 1 → skip,在收官報告寫
skipped: diary-gate BLOCK(理由)。exit 0 → 再過 DIARY-PIPELINE §Stage 0b 四個家路由:這段反芻如果是「做了 X 順帶注意到 Y」→ 留在 memory 的 Beat 5;「下次要記得 Z」→ LESSONS-INBOX;「這個 pattern 又遇到一次」→ 去 bump 既有條目;「跨好幾天才看得出的形狀」→ weekly-report。四個都不是,而且我對這件事的理解跟今天早上不一樣了,才寫日記。routine session 預設 skip(§Stage 0c):要寫才需要理由。純空場、純對賬全綠、純「今天也是零」一律不寫——那種反芻寫進 memory Beat 5 就夠了。
c.
/twmd-evolve— 數據驅動內容進化掃描(前瞻)觸發
Skill twmd-evolve。evolve pipeline 會:- 讀 EVOLVE-PIPELINE.md
- 跑 GA4 + SC 數據掃描
- 找出最有潛力的下次 rewrite candidate
- 寫進 ARTICLE-INBOX
注意:evolve 是前瞻 next-session 的工具,跟 memory/diary 是回顧不同。如果 session 沒有 ship 內容(純診斷 / 純 review),可 skip evolve。
-
Step 3: 最終 finale 報告(給觀察者一份完整收官 summary):
## 🧬 Session Finale Report — {session-id} ### Ship 統計 - Commits: N (hash list) - Files changed: M - New articles: ... (含 category/slug) - Babel 同步: en/ja/ko/es/fr 狀態 - Spore: # 篇 + 平台 ### 完成的 sub-skill - ✅ /twmd-memory → memory/{session-id}.md - ✅ /twmd-diary → diary/{session-id}.md (or skipped reason) - ✅ /twmd-evolve → ARTICLE-INBOX append (or skipped reason) ### Handoff 三態 (給下個 session) - [ ] pending: ... - ⏳ blocked: ... - [x] retired: ... ### 觀察者 callout / LESSONS-INBOX 候選 - ... (本 session 累積的教訓 candidates) 🧬 -
鐵律:
- 三 sub-skill 順序固定:memory → diary → evolve(先記錄、再反芻、再前瞻)
- chain 期間 git 操作謹慎:sub-skill 各自會 commit,finale 不要 batch reset
- finale 本身不 commit:每個 sub-skill 自己處理 commit + push。finale 只 orchestrate + 報告
- 觀察者中斷可重跑:finale 可重複觸發,sub-skill idempotent(已寫的 memory/diary 會 detect 不重寫)
- 「這次什麼都沒 ship」場景:純診斷 session 仍可跑 finale — memory 寫「本 session 純 review」,diary 預設 skip(沒動到世界的 session,反芻寫進 memory Beat 5 就夠;v2.5 前這裡寫「diary 寫反芻」,是 routine 每天產日記的來源之一),evolve 跳過
故意精簡。MEMORY-PIPELINE / DIARY-PIPELINE / EVOLVE-PIPELINE 各自 canonical 細節在 docs/pipelines/,finale 只是 orchestrator。
Signals
- GitHub stars
- 1k
- Forks
- 185
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
twmd-finale- Source
- github.com/frank890417/taiwan-md