Slash command
Also: 斜槓指令 · 斜線指令 · 自訂指令 · slash 指令
A short /command typed by the user that expands into a predefined workflow — triggered by a human, never selected by the model.
When you will meet it
You need one the day you notice yourself repeating the same three-minute briefing to the agent every morning. Confusing commands with skills picks the wrong mechanism: a skill is for the model to select on its own; a slash command is an explicit human order. Use the command for deterministic triggering, the skill for automatic adoption at the right moment.
An analogy
Like a text shortcut on your phone: typing @@ expands to your full address. You never retype the address, but you must type @@. A slash command is that shortcut for agents — except it expands into a whole workflow.
Minimal example
你輸入： /dd 0700
展開成（示意）：
「對股票代號 0700 執行完整盡職調查：
1. 用 hk-stock-dd 技能收集近三年財報
2. 分析股權結構與關聯交易
3. 輸出報告到 reports/ 目錄」
對照技能：你不打 /dd、只說「幫我看看這家公司」時，
用不用技能、用哪一個，由模型自己判斷。Note the split: slash commands solve deterministic triggering; skills solve method reuse. An expanded command can still invoke skills — they complement each other.
What people get wrong
- Assuming slash commands are smarter than natural language. They add no intelligence; they expand a fixed brief verbatim — the value is determinism and saved typing, not capability.
- Turning everything into slash commands and expecting the model to trigger them on its own. It will not — auto-triggering is the skill's job; a command only recognises the line the user types.