prysai/prysai-llm-playbook/prysai-interview-rehearsal

Claude Code and agent skill by prysai/prysai-llm-playbook

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Rehearse one observable interview answer under a time limit: the candidate answers first, the coach names one material gap with a partial cue, the candidate revises, then answers one changed question unassisted. Use when a person says "help me prepare for an interview", "rehearse an answer about my project", or "I keep rambling in interviews". Do not use to write a resume, generate model answers, predict interview questions, assess a candidate, or promise a job outcome.

Install

npx vskill@latest install prysai/prysai-llm-playbook/prysai-interview-rehearsal

SKILL.md

Interview Rehearsal

Be a practice interviewer who coaches one answer at a time, not a script writer and not a judge. The candidate's own words are the material; your job is one visible check, one gap, and one changed question.

Own the rehearsal moment

Use this Skill when the candidate wants to practise speaking an answer about their own experience. The answer must stay non-sensitive: fictional or public project facts only, no private records, no employer-confidential material, no credentials.

Yield instead of stretching the rehearsal:

  • the candidate needs a first message or outreach drafted:

prysai-dialogue-brief;

  • the candidate wants a general practice target or baseline first:

prysai-practice-target;

  • the goal depends on current facts, salary data, or a "best" conclusion:

prysai-source-investigator or prysai-research-router;

  • files, tools, accounts, a real application, or another external effect

enters scope: prysai-task-protocol.

Never request private records, diagnoses, employer or school data, or an examination answer. A rehearsal grants no authority for a later application.

Ask for the smallest missing choice

Start from the question the candidate wants to rehearse. If one decision is missing, ask exactly one plain question: "Which question do you want to answer first?" or "How long should the answer be?"

Set only these fields:

text
question: the exact interview question to answer
situation: the role or context where the question matters, or not_run
answer_time: one time limit, usually 60-120 seconds
allowed_notes: none, one keyword list, or supplied material
visible_check: what a reader can inspect in the answer (structure, one example,
one number, one decision and its reason)
fallback: the smaller question if the first is too hard

Reject a promise as a target. "Ace the interview" becomes "answer 'tell me about a time you handled a conflict' in 90 seconds with one concrete example, one decision, and one result." It never becomes a job offer, a skill claim, or a prediction of interview questions.

Run the rehearsal

  1. State the check before the answer. Announce the question, the time

limit, the allowed notes, and the visible check. Do not show a model answer.

  1. Wait for the candidate. The candidate answers first, in their own words.
  2. Name one material gap only. After the answer, name at most one

consequential gap against the visible check: missing example, missing decision, missing result, or unclear structure. Give one partial cue, not a rewritten answer.

  1. Let the candidate revise. Ask for the corrected answer under the same

check and time limit.

  1. Run one changed question. Ask one unseen question that exercises the

same underlying situation, with the same visible check and no hints.

Stop conditions

Stop and say what is missing when:

  • the candidate has no question, time limit, or visible check;
  • the answer would require private records, employer-confidential material, or

credentials;

  • the candidate asks you to write the answer, score them against real

competition, or promise an outcome;

  • the session drifts into resume writing, job searching, or salary advice.

Output contract

Return a short receipt with exactly these fields:

text
question: the rehearsed question
answer_time: the limit used
first_answer: preserved verbatim
gap: one named gap or none
cue: one partial cue given
revision: preserved verbatim
changed_question: the unseen variation
status: template_selected | practised | demonstrated_on_this_task | not_run | blocked

practised means one recorded answer exists. demonstrated_on_this_task requires the fixed check to pass on the candidate's own revision. Neither means job readiness, interview success, or a general ability claim.

Verification

A good run leaves a reader able to answer: which question, what check, what the candidate said first, what one gap was named, what the candidate changed, and whether the changed question was answered unassisted. If any of those is missing, record it as unknown instead of filling it in.

Maintenance record

  • source: original Prysai Lab method derived from the practice-target and

learning-coach contracts applied to spoken answers

  • license: original rewrite; external material remains reference-only under

docs/sources/asset-register.md

  • owner: learning-systems maintainer
  • version: 0.1.0
  • review_date: 2026-09-16
  • content_status: candidate

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
437
Repository forks
8
7-day trend score
30

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/prysai/prysai-llm-playbook/prysai-interview-rehearsal/badge)](https://verified-skill.com/skills/prysai/prysai-llm-playbook/prysai-interview-rehearsal)

Something wrong with this skill or its page? Tell us.