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。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/meridian-online/orbit --skill discovery命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Audit AC-to-test traceability — find untested ACs, orphaned test prefixes, and coverage gaps
Write a feature card — capture who needs what, why, and expected behaviours as scenarios
Extract capability cards from source material — files, directories, or a whole project
基于 SOC 职业分类
正在显示 SKILL.md
| name | discovery |
| description | Explore a vague idea through Q&A — capture what good looks like as an interview record for /orb:spec |
| argument-hint | [topic] |
| allowed-tools | AskUserQuestion Read Edit Write Glob Grep |
Explore a vague idea or requirement through questioning. Use this when there's no card yet, or the card is thin (< 3 scenarios). The outcome is clarity — what good looks like, captured as an interview record ready for /orb:spec.
/orb:discovery [topic]
If a well-formed card exists (≥ 3 scenarios), use /orb:tabletop instead.
.orbit/cards/ for any related feature cards. If a thin card exists, read it as starting context..orbit/specs/YYYY-MM-DD-<topic-slug>/ (create if needed)Before asking questions, run a keyword scan (see /orb:keyword-scan) against .orbit/specs/ and .orbit/choices/ using terms from the topic. If prior specs or decisions already explored this area, front-load what they found — the discovery session should build on existing knowledge, not rediscover it.
Adopt the interviewer role (see /orb:interviewer for the full persona and the decision-level gate).
The author's job is to define what good looks like. Discovery questions stay at the intent level — goals, users, constraints, success criteria. Implementation questions are recorded as notes, not asked.
Target: 6–10 questions covering:
For each question:
Present the question using AskUserQuestion with contextually relevant suggested answers:
Record the Q&A pair in your working notes
After each answer, target the biggest remaining source of ambiguity
After every 3-4 questions, assess clarity across three dimensions:
If all three are clear (ambiguity ≤ 0.2), suggest wrapping up.
During discovery, choices between approaches will surface. When a clear choice is made:
Save the Q&A as: .orbit/specs/YYYY-MM-DD-<topic-slug>/interview.md
# Discovery: <Topic>
**Date:** YYYY-MM-DD
**Interviewer:** <agent name>
**Card:** .orbit/cards/NNNN-slug.yaml (if applicable)
**Mode:** discovery
---
## Context
<Initial topic/idea description>
## Q&A
### Q1: <Short label>
**Q:** <question>
**A:** <answer>
[...]
---
## Summary
### Goal
<Extracted goal statement>
### Constraints
- <constraint 1>
### Success Criteria
- <criterion 1>
### Decisions Surfaced
- <choice made>: chose X over Y because Z (→ .orbit/choices/NNNN if recorded)
### Implementation Notes
- < — >
Next step: /orb:spec to generate a structured specification from this discovery.