一键导入
archi-interview-protocol
Inline helper for structured multiple-choice interviews. Protocol-invoked only; must return user-facing questions to the main conversation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Inline helper for structured multiple-choice interviews. Protocol-invoked only; must return user-facing questions to the main conversation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Fetch minimal Architext documentation context after an Intent Card. Must run in isolated context/subagent when available. Auto-invoked by 00_system Front Pipeline.
Normalize a user request into an Architext Intent Card before command routing. Must run in isolated context/subagent when available. Auto-invoked by 00_system Front Pipeline.
Reconcile stale Architext task status from evidence. Must run in isolated context/subagent when available. Protocol-invoked only; do not auto-trigger from casual user requests.
Fetch minimal Architext documentation context after an Intent Card. Must run in isolated context/subagent when available. Auto-invoked by 00_system Front Pipeline.
Normalize a user request into an Architext Intent Card before command routing. Must run in isolated context/subagent when available. Auto-invoked by 00_system Front Pipeline.
Reconcile stale Architext task status from evidence. Must run in isolated context/subagent when available. Protocol-invoked only; do not auto-trigger from casual user requests.
| name | archi-interview-protocol |
| description | Inline helper for structured multiple-choice interviews. Protocol-invoked only; must return user-facing questions to the main conversation. |
/archi.* protocols.Multiple-choice first. Do not open-ended ask (e.g. "What database do you want?"). Each question gives 3-5 options + [Z] Custom, AI recommends one, user confirms or chooses differently. Goal: User without domain knowledge can make reasonable decisions.
gaps, do not self-expand topicscontext do not re-ask, directly adopt[Z] Custom (please describe) fallback[Recommended], reason must combine context, do not mechanically apply defaultAI+ / AI- write complete sentences from AI Agent execution perspective; each option must have tradeoffs, do not write "none"max_questions### Supplementary Confirmation
**[Q<n>] Question Title**
> One sentence explaining why this information is needed
| ID | Option | Description | AI+ | AI- |
|:---|:---|:---|:---|:---|
| A [Recommended] | Option name | 2-3 sentence description | Complete sentence | Complete sentence |
| B | ... | ... | ... | ... |
| Z | Custom | (please describe) | - | - |
(Multiple questions repeat above Q-table)
**INPUT**: `Q1 answer | Q2 answer | ...` (Between questions use `|`; Single question multiple choices use space, e.g. `A B`)
| Error | Rule Violated | Fix |
|---|---|---|
| `A | PostgreSQL` | #6 One-word description |
AI+: Good performance | #7 Not complete sentence | AI+: Structured Schema lets AI directly infer CRUD from types, reducing guessing |
AI-: None | #7 Avoiding | AI-: Migration scripts need sync maintenance as Schema evolves, AI easily misses field changes |
| Asking about tech stack Brief already stated | #2 | Skip, directly adopt |