| name | ReadAware Onboarding |
| description | 访谈读者本人、建立或更新阅读偏好档案（profile）。当用户第一次使用 readaware、想设置/修改"共读时该怎么跟我说话"、说"重新做下我的阅读偏好"时触发。Interview the reader and write/update their reading profile that drives how read behaves. |
| allowed-tools | Bash, Read, Write, AskUserQuestion |
readaware:onboarding — build a base map for this reader
You're interviewing the person using this skill right now, to produce a
~/.claude/readaware/profile.md. readaware:read loads it on every co-reading session to
decide what tone, depth, and format to speak to this person with.
Hard rules
- Don't assume you already know them. You don't know who they are, their age, what
they've read, or why they're reading this book — ask it out. Don't apply any template persona.
- This is an interview, not a form. The worst possible outcome is a questionnaire like
"What's your reading preference? □ Depth □ Fun". Ask specific, hooked, follow-up
questions — pull threads from their answers rather than mechanically reciting a list.
- Profile quality = the quality of every future session. A vague "likes deep analysis"
is useless; force out executable preferences ("quote the original when giving examples,
don't paraphrase"; "I can't get through dense walls of text").
- Default to the user's language throughout (Chinese first, unless they use another).
How to interview
Use AskUserQuestion to throw 1–3 good questions at a time, and dig deeper based on the
answers. Cover these dimensions (no need to ask each one literally — draw them out naturally):
- Who the reader is (the reading-relevant part) — background, knowledge terrain, why
this book, what they want out of it. Don't mine privacy; mine what's needed to
understand how they read.
- How to talk to them — tone (academic / peer conversation / sharp), depth, length
preference, layout preference (can they take long passages, do they want the conclusion
first then the unpacking), and the kind of expression they hate most (AI-speak? a
teacherly tone? empty phrases?).
- What kind of co-reading they want — should you interpret, ask Socratic questions,
debate them, play devil's advocate? Should you proactively connect motifs/structure, or
only answer what's asked?
- The lenses they return to — which angles they default to (sociology? philosophy?
psychology? history? pure close reading of the text?).
- Landmines — what makes them close the tab instantly.
Stop once you can draw a clear base map. Don't go on forever.
Output
Organize the interview into ~/.claude/readaware/profile.md, clearly structured as
operating instructions for read rather than prose. Suggested sections:
# Reader profile (readaware profile)
## Who the reader is
## How to talk to me (tone · depth · length · layout)
## The kind of co-reading I want
## Lenses and motifs I return to
## Landmines (what not to do)
Run mkdir -p ~/.claude/readaware before writing. After writing, read the profile back to
the user and ask them to revise it — this is their base map, they need to own it.
Updating an existing profile
If ~/.claude/readaware/profile.md already exists: read it back to the user first, ask which
part they want to change, and update incrementally — don't silently overwrite from
scratch. That's the whole point of this being standalone and re-runnable anytime.
