Starts a mock interview session with configurable mode (guided, random, adaptive), stage (recruiter, hiring manager, technical, panel, executive, final), and vibe (supportive through bored). If mode or vibe are omitted, the system selects defaults and announces them. Delegates to interview-coach; optional mcp-voice MCP TTS/STT when tools are present. Also invocable via /career-navigator:mock-interview.
{user_dir} and gateSame as prep-interview: require CareerNavigator/profile.md, ExperienceLibrary.json, and tracker.json unless the user only wants a generic mock with pasted JD (then still need profile + EL minimum; create minimal context note if tracker empty).
| Parameter | Values |
|---|---|
| mock_mode | guided | random | adaptive |
| interview_stage | recruiter | hiring_manager | technical | panel | executive | final |
| vibe | supportive | | | | |
neutralchallengingantagonisticbored| target | application_id or company + role (optional for generic practice) |
The model must choose a concrete mock_mode and vibe before starting—never leave them implicit in the opening line.
mock_mode (if user omitted): use adaptive (default for learning and feedback). If the user asked for “surprise me,” “mix it up,” or equivalent, pick random at random and state that you did.vibe (if user omitted): use neutral. If the user asked for “surprise” / “mix it up,” pick one vibe at random from the five allowed values and state it.interview_stage (if user omitted): infer from tracker.json (next likely stage for the active application—e.g. phone_screen / recruiter → recruiter; otherwise hiring_manager). If no application context, default hiring_manager.Selected: mode=adaptive, vibe=neutral, stage=hiring_manager (defaults — say if you want different).Recruiter practice: When the user asks to practice for a recruiter or phone screen, set interview_stage to recruiter.
Discover session tools: if speak (TTS) or listen (STT) from the mcp-voice extension are present, prefer them for mock practice per agents/interview-coach/AGENT.md. Otherwise text-only. Never require audio.
Read and follow agents/interview-coach/AGENT.md with:
mode: mockmock_mode, vibe, interview_stageRun the mock turn-by-turn in chat until the user ends or you reach a natural wrap-up per the agent.
If the user wants a log: append a short applications[].notes entry (no [prep] required) e.g. [mock] with date and stage—only when they ask or when it aids follow-up.