English · 中文
Two things that work together, and work separately.
- A skill for Claude — builds the two research assets that make interview preparation actually work: an employer dossier with a defensible why them, and a STAR story bank with the evidence behind why me. Available in English and 中文.
- The interviewer (
interviewer/index.html) — a single HTML file you open in Chrome to practise against a real clock, with video, transcript, and a debrief your AI coach can mark up.
The premise behind both: interview outcomes turn on fit, and an AI can only say something useful about fit if it can read real material about you and about them. Ask a model to "run a mock interview" cold and it produces answers that fit any candidate and any employer — which is exactly what loses interviews. So the skill produces material first, and the interviewer measures your performance against it.
git clone https://github.com/Frikasong/mock-interviewer.git
cd mock-interviewer
./install.sh en # English edition; ./install.sh for 中文The installer copies the skill into ~/.claude/skills and warns you if the other
language edition is already sitting there — the two compete for the same requests,
so install one. Restart Claude Code afterwards.
Claude.ai — download the .skill file from Releases and upload it
under Settings → Capabilities → Skills.
Neither — paste skills/interview-prep/SKILL.md into a conversation. It works; you
just re-paste it each time.
Then describe your situation rather than naming the skill: "I'm interviewing at Tidewater Outfitters for a product marketing role next Thursday, here's the posting." It takes it from there.
Download interviewer/index.html and open it
in Chrome. That is the entire installation procedure — no Node, no server, no API key, no
account. See its README for the question format and
browser support.
skill interviewer
───── ───────────
step 1 employer research ──┐
dossier, why-them │
question-bank.md ───┼──────────► loads directly
│ (### question + fields)
step 2 self research ──┤
story cards S01… ───┼──────────► pasted as "what you prepared"
selling points │
│
step 3 mark the debrief ◄──┴──◄──────── debrief.md comes back:
answered badly → back to the bank
new story surfaced → new card
Three sessions in, your question bank is more accurate than any published interview report, because it grew around your own weak points rather than someone else's.
mock-interviewer/
├── build.py builds every release artifact
├── DEPLOY.md release process
├── skills/
│ ├── interview-prep/ English skill
│ │ ├── SKILL.md
│ │ ├── references/ employer research · story bank · mock & debrief
│ │ └── assets/ blank templates for every output
│ └── interview-prep-zh/ 中文版，结构相同
├── interviewer/
│ ├── index.html the whole tool
│ ├── README.md
│ └── scripts/ five ready-made question banks
├── install.sh copies the skill into ~/.claude/skills
├── docs/ release notes for other channels
└── LICENSE
A few decisions worth stating, since they're the ones people usually ask about:
The skill never writes your experience for you. It interrogates, structures, and sharpens; the facts have to come from you. A fabricated story survives exactly one follow-up question, and the failure mode is worse than having no story.
The interviewer contains no AI. No key, no cost, no recording of yours passing through anyone's inference. The debrief is a markdown file designed to be pasted into whatever assistant you already use, with the coaching prompt written at the top.
Nothing is uploaded, ever. Video, audio and transcript live in the browser tab and are discarded when you close it unless you download them first. For a file full of your own unrehearsed interview answers, that is the correct default.
MIT — see LICENSE.