Interview knowledge format

Skill XiaoChu-1208/interview-assistant-CLI/skills/interview-knowledge-format

Real-time interview copilot CLI. Hears the question and surfaces the answer you already prepared from your local Markdown knowledge base — verbatim, in milliseconds. Instant recall instead of LLM hallucination.

Install

npx -y skills add XiaoChu-1208/interview-assistant-CLI --skill interview-knowledge-format

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Convert résumé / job descriptions / past interview notes into Markdown Q&A files in the format Interview Assistant's RAG expects (sections + Q/A pairs + optional Unicode tree blocks). Use when the user wants to populate `knowledge/` for the first time, ingest a new role, or refactor messy notes.

SKILL.md

2.7 KB, as published. Nobody here has run it

interview-knowledge-format

A Skill for both AI editors (Cursor / Claude Code / Codex) and the Interview Assistant runtime. The runtime injects the format spec below into the chat system prompt so the LLM stays aligned even mid-session.

When you generate or edit knowledge-base content, follow this exact format.

File rules

One topic per file. Filename = lowercase-snake, e.g. about_me.md, pm_methodology.md, data_analytics.md. Drafts use prefix draft_.
UTF-8, LF line endings. No BOM.
Top of file: # <Topic> (H1) + a one-line tagline if helpful.

Section rules

Each retrievable unit MUST start with ## (H2). The H2 title IS the question. Multi-question files are encouraged.

## How do you handle conflict on a cross-functional team?

**Q:** How do you handle conflict on a cross-functional team?

**A:**

<concise STAR answer in the user's first person, 4-8 sentences>

> Bullet 1: opening line you'd actually say out loud
> Bullet 2: concrete data point
> Bullet 3: result with metric

Optional Unicode tree

If the user wants a visual cheat-sheet rendered in the terminal, append a fenced tree block at the END of the section:

```tree
▶ Conflict resolution
├─ ① ack the friction
├─ ② re-anchor to user metric
└─ ③ propose an A/B test
```

The runtime renders this with color. Keep ≤ 6 lines, ≤ 60 chars per line.

Hallucination guard

Quote sources verbatim where possible.
If source data is missing, write [需补充] (Chinese) or [TBD] (English) — never invent numbers.
Never substitute company names or metrics.

How to use this skill

From an AI editor

"Use the interview-knowledge-format skill. Convert my résumé into one Q&A file per role under knowledge/."

The AI will:

Read the source.
Group into topics.
Emit one MD file per topic in the format above.
Mark each with draft_ prefix until you review.

From the runtime

The Interview Assistant auto-loads this skill at startup; the spec above is appended to the LLM system prompt so all generated answers stay format-aware.

Templates

See templates/example-qa.md for a full filled-in example.