| name | alt-import |
| description | Parse an Exam Radar (OPTIMETA Alt plugin) export and fold its lecture-emphasis exam-probability signal into the PAIDEIA course index — write course-index/radar.md, annotate course-index/coverage.md with a lecture-emphasis column and divergence flags, and seed a gold-zone weakmap. Invoked by /paideia:alt. The export form is fixed (exam-radar:v1 marker). |
alt-import
Exam Radar is OPTIMETA's Alt plugin. From lecture-recording transcripts it extracts the topics a professor verbally emphasized, ranks them by exam probability, and lets the user triage each into one of three zones. Its 복사 button emits a fixed markdown form. This skill ingests that form.
PAIDEIA already has one exam-probability signal — HW density (course-index/coverage.md). Exam Radar adds a second, independent one — lecture emphasis. The two corroborate where they agree and expose blind spots where they diverge.
Premise (do not break). HW density remains the primary Exam tier. Lecture emphasis is layered on as annotation and a second opinion — it is surfaced, never substituted. As with a user-declared weak zone, a single lecture signal does not auto-upgrade an HW-based tier (mirror course-builder's rule). What it does is flag divergences for the user to judge.
1. The export format (exam-radar:v1)
# Exam Radar 작전 — <course>
<!-- exam-radar:v1 source=alt -->
- 코스: <course>
- 시험까지: <D-N>
- 토픽: 총 <N>개 (골드존 <G> · 버려도 안전 <D>)
- 버려도 안전 비중: 전체의 <P>%
## 지금 할 것 — 골드존 (시험확률 높음 · 아직 약함)
1. <topic> · 시험확률 <p>%[ · 🎙]
...
## 이미 다진 것 (잘 알거나 시험에 덜 나옴)
- <topic> · 시험확률 <p>%
...
## 버려도 안전 (안 해도 되는 것)
- <topic> · 시험확률 <p>%
...
Parse rules:
- Detect by the
<!-- exam-radar:v1 marker. Parse the vN integer; if > 1, warn and parse the v1 fields best-effort (ignore unknown ones).
- Meta: read
- 코스:, - 시험까지: (D-N), and the count line.
- Zones: three
## headings →
지금 할 것 — 골드존 → zone gold (high exam-prob, low self-confidence).이미 다진 것 → zone strong (already known or low exam-prob).버려도 안전 → zone skip (safe to drop).
- Each topic line:
<name> · 시험확률 <p>%, optionally · 🎙 (the professor verbally stressed it). The leading 1. / - is just list markup. Parse name, integer p (0–100), and the 🎙 flag.
- Be lenient on whitespace and the leading marker. A zone with
(없음) / (아직 없음 …) is empty.
2. Write course-index/radar.md (canonical store)
This mirrors how coverage.md stores the HW signal. Overwrite on re-import.
<!-- SOURCE: Exam Radar (Alt), exam-radar:v1, course=<course>, <D-N>, imported <YYYY-MM-DD> -->
# Lecture-emphasis signal — <course>
Imported from Exam Radar. Exam probability here is **lecture emphasis** (professor's spoken stress + repetition across recordings), independent of HW density in `coverage.md`.
| Topic | Exam prob | Zone | Lecture signal |
|---|---|---|---|
| <topic> | <p>% | gold | 🎙 |
| <topic> | <p>% | strong | — |
| <topic> | <p>% | skip | — |
...one row per topic, exam-prob descending within each zone, gold → strong → skip.
## Now (gold zone)
High exam probability, still weak — drill these first:
- <topic> (<p>%)[ 🎙]
## Safe to drop
Low lecture emphasis — reference only:
- <topic> (<p>%)
If course-index/ doesn't exist yet, create it. radar.md stands on its own even before /paideia:analyze has run.
3. Merge into course-index/coverage.md (if it exists)
If coverage.md is missing, skip this step and tell the user to run /paideia:analyze first — radar.md already captured the import.
Otherwise:
-
Map each Exam Radar topic to a reverse-map section (§) by title match — case- and spacing-insensitive, substring allowed (e.g. "Gram-Schmidt" ↔ "§3.2 Gram-Schmidt orthogonalization"). Keep the best match; leave the rest unmatched.
-
Add/refresh a Lecture emphasis column on the reverse-map table (§ | Title | HW coverage | Exam tier | Lecture emphasis). Value from the topic's exam-prob:
🎙🎙 — gold zone or ≥ 70%🎙 — 40–69%· — < 40%, or no Exam Radar topic mapped to that §
-
Do not change the Exam tier. It is HW-derived and stays. Lecture emphasis is the new column only.
-
Append a divergence section — this is the payoff of having two signals:
## Lecture vs HW — divergences (judge these)
### 🎙 Stressed in lecture, but no HW
§/topic the professor emphasized (🎙🎙) yet `coverage.md` marks ⚪ Low-risk:
- <§ or topic> — verbal-only exam point? decide; if it matters, `/paideia:derive` or `/paideia:quiz` it.
### HW-dense, but quiet in lecture
§ marked 🔥🔥/🔥 with `·` lecture emphasis:
- <§> — quietly important; the professor tests it without spending lecture time on it.
Respect the premise: the ⚪→ "judge this" line is a prompt for the user, not an automatic tier upgrade.
-
Unmatched topics — append so nothing is lost:
## From Exam Radar (no HW section match)
- <topic> · <p>%[ · 🎙]
-
Drill priority — in coverage.md's "Recommended drill priority", use lecture emphasis as a booster/tie-breaker (a 🎙🎙 + thin/blind item ranks above an equal one without emphasis), without reordering across HW tiers.
Re-runs replace the Lecture emphasis column and the Lecture vs HW / From Exam Radar sections in place — never duplicate them.
4. Seed a gold-zone weakmap
The gold zone = high exam probability + low self-confidence = a weakness the user effectively declared, corroborated by lecture emphasis. Treat it exactly like /paideia:weakmap Case B (user-declared weaknesses).
- Write a new
weakmap/weakmap_<YYYY-MM-DD_HHmm>.md (never overwrite — preserve history).
- Use the weakmap report format (see
commands/weakmap.md). Put gold-zone topics under ## User-declared weaknesses, each tagged (from Exam Radar gold zone), mapped to related §/Pk (via course-index/), with a recommended drill (/paideia:blind / /paideia:quiz / /paideia:derive).
- Also run the weakmap Case A latest-error snapshot so the report stays consistent. If
errors/log.md and course-index/ are absent, write a minimal report (gold zone only) and note that /paideia:analyze will enrich it.
5. Notes
- Lecture-emphasis exam-prob and HW-density exam-prob are different axes; never average them into one number. Keep them as separate columns/files so the user can see both and reason about divergence.
- Exam Radar topics are lecture-derived, so they overlap heavily with
summary.md sections — but the user may have edited section numbering. Always do best-effort matching and surface the unmatched rather than forcing a map.
