| name | interview-state-tool |
| description | Use when inspecting simulated interview Agent short-term session state, job-resume binding, question plan progress, answer signals, difficulty, follow-up strategy, or remaining focus. |
Interview State Tool
Tool Contract
Use this skill to inspect interview_sessions.report.agent_state. The state story is JobProfile -> CandidateProfile -> QuestionPlan -> AskedQuestions -> AnswerSignals -> EvaluationState -> FollowupStrategy -> SummaryReport.
Current memory boundary:
- Interview assistant uses current-session short-term memory only.
- It does not read AI assistant file memory.
- It does not write complex long-term
assistant_memories records.
- Long sessions may compress older turns into
agent_state.session_summary inside the same interview session.
Script Usage
Run inside the API container:
docker compose exec api python /app/skills/interview-state-tool/scripts/inspect_interview_state.py --email admin@example.com
Inputs: optional --session-id, optional --user-id or --email. Without --session-id, the latest user session is inspected.
Output Contract
The script emits compact JSON: available, session_id, job_id, resume_id, mode, message_count, has_agent_state, difficulty, remaining_focus, last_followup_strategy, asked_count, and state_keys. It does not expose hidden reasoning or full message transcripts.
Answer Synthesis
Explain where the interview currently is, what focus remains, and why the next follow-up may change. If no state exists, tell the user to start or resume a bound interview session.
Validation
python skills/interview-state-tool/scripts/inspect_interview_state.py --self-test
python skills/interview-state-tool/scripts/inspect_interview_state.py --help
docker compose exec api python /app/skills/interview-state-tool/scripts/inspect_interview_state.py --email admin@example.com
docker compose exec api pytest tests/interview/test_interview_short_memory.py -q