databy frank890417
🧬 Taiwan.md — Refresh Dashboard Data v3.0
frank890417/twmd-refresh
Refresh Taiwan.md dashboard data — full ground truth抓取與更新 via canonical DATA-REFRESH-PIPELINE (git pull + 三源感知 + prebuild dashboard JSON 全套 regen + GitHub stats). TRIGGER when: user says "資料更新", "refresh", "跑 refresh-data", "三源感知", "dashboard 更新", "dashboard 資料抓取", "dashboard 怪怪的".
What it changes
Passed 12 of 15 test cases with the skill, and 2 of 15 without it.
Without skill14%
With skill77%
Checks it passed that the model alone missed
- Specifies recording handoff three-state information
- Specifies including Beat 5 reflection in the output
- Identifies the exact path scripts/tools/refresh-data.sh
and 13 more in the full report
15 fixed2 passed either way4 still failing1 broke
- Pass-rate lift
- +64 pp
- Tokens
- -6%
- Agent turns
- 0%
- Response time
- -46%
Model gemini-3.6-flashMeasured 8/16/2026v117 of 22 cases passed with skill · 15 comparable
See it in action
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
PromptWhen running the finale chain after pipeline completion, what acknowledgment line must be included at the top of the memory file?
What was checked, without and with the skill| Checked | Without skill | With skill |
|---|
| Requires including the BECOME ACK line in the memory file | Missed | Passed |
|---|
What's inside
The instructions your model reads when the skill loads.
SKILL.md48 linessingle file
🧬 Taiwan.md — Refresh Dashboard Data v3.0
🚨 STRICT BECOME GATE — 第一動作不可省略
Before anything else：跑 /twmd-become micro 完整走 BECOME_TAIWANMD.md Step 0-9。Micro mode self-test 7 題全過才能進 Stage 1。
✅ BECOME ack: mode=micro / 8 organ 最低=<即時 consciousness-snapshot.sh> / Q14 cross-session continuity=PASS
Stage 1: 跑 pipeline
bash scripts/tools/refresh-data.sh
步驟表不在本 skill 複寫 — canonical：DATA-REFRESH-PIPELINE.md §一鍵執行（步數與編號以 refresh-data.sh 頭註為準）。
Rider：scheduler live-state dump（session 層步驟，bash 進不了 MCP store——2026-07-05 S1 根治）：
- 呼叫
mcp__scheduled-tasks__list_scheduled_tasks，raw JSON 存暫存檔
python3 scripts/tools/routine-live-normalize.py <raw.json> --session <session-id>（自動過濾私人 routine，只留 twmd-/taiwanmd-）docs/semiont/routine-live-state.json 跟著 refresh commit 進 git — routine-sync-check.py v3 第三層比對靠它
2026-07-05 前本 skill 曾複寫 14 步全表——違反 ROUTINE-PROMPT-CONTRACT「殼層禁複寫 SOP」，且複寫版一度比 canonical 還新（SSOT 失守症狀，dna-audit §S5 反向案例）。改 pointer 後：修 pipeline = 改一處。
Stage 2: 報告 + Step 11 freshness gate handling
Step 11 抓到 stale dashboard JSON → 走 DATA-REFRESH-PIPELINE §catch ≠ fix 鐵律（第 2 次連續 catch 同一 stale 必須當 cycle wire fix，不 spawn chip——鐵律全文與誕生背景在 canonical，本殼只留觸發指令）。
Stage 3: 收官
/twmd-finale chain → memory file 必含：
Track record
Benchmark
12 / 15
test cases passed, vs 2 / 15 without
Adoption
- Users
- 0
- Times used
- 0
Safety and source details
SkillSafety
✓ 3 of 3 passed✓Code scan15 of 15 checks clean
0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026
View 15 scanner checks
live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing
✓IntentSafe
LLM judge · does the skill do what it claims
✓ContentClean
LLM judge · policy & harmful-content review
Source
ImportedImported 8/3/2026Version v1