SKILL.md
推導工作流 Skill
核心原則
SymPy-MCP 做計算，NSForge 記錄知識！
每步都要顯示給用戶看！ 使用：
-derivation_show()(NSForge) - 顯示當前推導狀態
-printlatexexpression()(SymPy-MCP) - 顯示計算結果
⚠️ 黃金法則：永遠向用戶展示公式！
❌ 錯誤：執行計算後直接下一步
✅ 正確：執行計算 → derivation_show() 或 print_latex_expression() → 等用戶確認 → 下一步工作流程
Phase 1: derivation_start(name, description)
↓
Phase 2: 循環 {
SymPy-MCP: intro_many → introduce_expression → 計算 → print_latex_expression
NSForge: derivation_record_step(expression, description, notes?)
NSForge: derivation_show() ← 🆕 顯示當前狀態！
NSForge: derivation_add_note(note, note_type?) # 可選
}
↓
Phase 3: derivation_complete(description, assumptions?, limitations?, references?)
derivation_show() ← 🆕 顯示最終結果！工具速查
| 階段 | MCP | 工具 | 用途 |
|---|---|---|---|
| 開始 | NSForge | derivation_start(name, description) |
建立會話 |
| 計算 | SymPy | intromany, introduceexpression, substitute_expression... |
符號計算 |
| 顯示 | SymPy | printlatexexpression |
⚠️ 必須！ |
| 記錄 | NSForge | derivationrecordstep(expr, desc, notes?, source?) |
記錄步驟+知識 |
| 🆕 顯示 | NSForge | derivationshow(format?, showsteps?) |
⚠️ 必須！顯示當前狀態 |
| 說明 | NSForge | derivationaddnote(note, note_type?) |
純文字洞見 |
| 完成 | NSForge | derivation_complete(...) |
存檔+元資料 |
note_type: assumption, limitation, observation, correction, clinical, physical
🆕 步驟 CRUD 操作
| 操作 | 工具 | 用途 |
|---|---|---|
| 📖 Read | derivationgetstep(step_number) |
查看單一步驟詳情 |
| ✏️ Update | derivationupdatestep(step_number, notes?, assumptions?, ...) |
更新步驟元資料 |
| 🗑️ Delete | derivationdeletestep(step_number) |
刪除最後一步 |
| ⏪ Rollback | derivationrollback(tostep) |
⚡ 回滾到指定步驟 |
| 📝 Insert | derivationinsertnote(afterstep, note, notetype?) |
在指定位置插入說明 |
Handoff：NSForge 做不到時
⚠️ Phase 2 後更新：Laplace/Fourier 變換已實作，無需 Handoff！
當需要 ODE/PDE、矩陣運算、聯立方程組：
# 1. 導出
result = derivation_export_for_sympy()
# → 返回 intro_many_command, current_expression
# 2. SymPy-MCP 計算
intro_many([...])
dsolve_ode(...) / solve_linear_system(...)
print_latex_expression(...)
# 3. 導入回 NSForge
derivation_import_from_sympy(
expression="...",
operation_performed="Solved ODE",
sympy_tool_used="dsolve_ode",
notes="...",
assumptions_used=[...],
limitations=[...]
)調用範例
# Phase 1
derivation_start("temp_mm", "Temperature-corrected Michaelis-Menten")
# Phase 2a: SymPy 計算
intro_many([{"name": "V_max", "assumptions": ["positive"]}, ...])
mm = introduce_expression("V_max * C / (K_m + C)")
print_latex_expression(mm)
# Phase 2b: NSForge 記錄
derivation_record_step(
expression="V_max * C / (K_m + C)",
description="Base Michaelis-Menten",
notes="假設溫度恆定"
)
# Phase 2c: 加入洞見
derivation_add_note("酵素在 >42°C 會變性", note_type="limitation")
# Phase 3
derivation_complete(
description="...",
assumptions=["Michaelis-Menten kinetics"],
limitations=["Valid for 32-42°C"],
tags=["enzyme", "temperature"]
)🆕 步驟 CRUD 範例
# 📖 Read: 查看第 11 步詳情
derivation_get_step(11)
→ {"step": {"description": "...", "notes": "...", "output_latex": "..."}}
# ✏️ Update: 更新第 11 步的說明
derivation_update_step(
step_number=11,
notes="此假設在高溫時不成立",
limitations=["Valid only for T < 42°C"]
)
# ⏪ Rollback: 發現第 11 步開始走錯，回滾到第 10 步
derivation_rollback(to_step=10)
→ {"deleted_count": 6, "deleted_steps": [11, 12, ...], "current_expression": "..."}
# 現在可以從第 10 步重新開始推導！
# 📝 Insert: 在步驟 5 和 6 之間插入說明
derivation_insert_note(
after_step=5,
note="此處假設達穩態",
note_type="assumption"
)