Coach Interview Prep End to End

What problem does this solve?

Coaches your full interview lifecycle as one adaptive system instead of treating each stage independently. It scores every answer across five dimensions, diagnoses root causes behind weak spots, tracks which stories work and which don't, learns from real outcomes, and cascades that learning to every future interview prep, practice round, and negotiation.

How does it work?

Once set up, the system maintains persistent coaching state across sessions. Run commands as you progress through your job search: `research` and `decode` for target evaluation, `prep` and `practice` for interview preparation, `round` for post-interview capture, `progress` to track patterns, `negotiate` for offers. Each command builds on prior context (your storybank, prior company research, previous interview scores, and lessons learned). Output: sustained improvement in interview performance measured across all five dimensions, with strategic direction that adapts as the system learns what works for you.

Included Skill Groups

What's the biggest win?

Every interview gets better because the system learns from the last one. Patterns surface before they become habits, stories that work stay sharp, and weak spots get targeted drills rather than generic coaching.

What should I know technically?

Requires Claude Code, Cursor, or any environment with file system access and paid Claude plan. Maintains all state in a persistent `coaching_state.md` file (gitignored by default). Optionally integrates with Minutes (silverstein/minutes) for auto-detecting interview transcripts from `~/meetings/`. Uses Bash for date verification (critical for scheduling accuracy). No external APIs or auth required. Everything runs locally.

What should I watch out for?

This system requires sustained engagement. Single commands won't move the needle. Best results come from running `kickoff` to establish baseline, building a storybank, running practice and mocks regularly, and capturing real interview outcomes via `round`. For quick one-off help, use `prep` or `resume` alone, but you'll miss the adaptive learning that makes the system powerful. Also: `coaching_state.md` contains sensitive personal and professional data (comp strategies, interview feedback, recruiter notes). Do not sync to shared cloud storage unencrypted.

Key Features