intervy-scorelisted

# intervy-score — Interview Evaluator ## Step 0 — Detect Session Type Read the conversation context and check for the presence of `[PANEL SETUP — internal]`. - If found → this is a multi-panel session. Read `multiple-panel-score.md` and follow it. Stop here. - If not found → this is a single-interviewer session. Continue to Step 1 below. --- ## Overview The candidate stopped here — by choice or because they ran out of time. That is completely fine. Evaluate only what they completed. Never mention or penalise skipped rounds. --- ## Step 1 — Collect Scores from Context Read the conversation and extract every score explicitly recorded by the interview skill. Build this table — mark only components that have a real score: | Component | Max | Score | Attempted | |---|---|---|---| | Project Bootstrap | 10 | | | | Round 1 Theory | 10 | | | | Round 1 Code | 10 | | | | Round 1.a Theory | 10 | | | | Round 1.b Theory | 10 | | | | Round 2 Theory | 10 | | | | Round 2 Code | 10 | | | | Round 3 Theory | 10 | | | | Round 3 Code | 10 | | | | Between-Round Probes | 10 | | | | Round 5 Theory | 10 | | | | Round 5 Code | 10 | | | Only rows with a recorded score count. Ignore the rest entirely. **Finding Between-Round Probe scores:** Search the conversation for lines matching `*[Probe scores: code=X/5 · language=Y/5]*`. Each such line is one probe invocation. Compute the score as: ``` per_probe = code + language (max 10 each) avg_probe = round(sum(per_probe) / count_of_prob