Run Full Interview Prep End to End

Runs your full interview search and interview prep as one adaptive system instead of jumping between generic AI help and fragmented tools. It scores every answer on five dimensions, diagnoses root causes behind weak spots, builds a storybank you retrieve under pressure, learns from your real interview results, and adapts its coaching to your specific patterns. Not a question bank. An adaptive system that gets sharper the more you use it.

Once installed, you run through four phases. 1. Setup phase: run kickoff with your resume, target roles, and interview timeline. The system builds your profile, assesses your starting point, and gives you a time-aware action plan. 2. Prep phase: run research and prep for specific companies, or run decode on job descriptions to assess fit before you apply. 3. Practice and mock phase: run practice drills (constraint ladders, pushback handling, panel simulations, stress tests) or full mock interviews in behavioral, system design, case study, or panel formats. 4. Post-interview phase: run round right after a real interview (with or without transcript) to capture impressions, score your performance, and update your coaching plan. All state saves to coaching_state.md automatically. Output: a comprehensive record of your progress, pattern trends, and a personalized next-action plan after every major command.

You practice against real scoring dimensions, see why weak answers actually fall short, and build pattern awareness across 3+ interviews. By interview 4-5, you're not drilling the same mistakes twice.

Requires Claude 3.5 Sonnet or higher (uses Claude Code API for file persistence). The coaching_state.md file is automatically gitignored to prevent accidental commits of sensitive data (storybank, feedback, comp strategy, transcript excerpts). Optional: install the minutes CLI tool to auto-detect recorded interview transcripts from ~/meetings/ and bypass manual transcript pasting. The skill loads command workflows on demand from the references/commands/ directory, so it can handle long deployments without exceeding token budgets per session.

Requires Claude Code, Cursor, or similar environment with file system access and a paid Claude tier. Long job searches can create large coaching_state.md files that consume significant context. Periodically compress closed interview loops to an archive file. Voice profile personalization (matching your writing style) requires you to share writing samples in the kickoff command. If you skip this step, output defaults to generic professional tone.