syllabus

---
name: syllabus
description: Use when a user shares a course syllabus and wants a supplementary reading list of recent peer-reviewed papers, grouped by topic, with plain-language summaries and discussion questions.
category: Research
version: 1.0.0
tools: []
---

# Syllabus — Course Supplementary Reading List

Build a professional supplementary reading list for an instructor or student, covering recent peer-reviewed papers per course section, delivered as formatted chat markdown.

## Agent Integrity Rules

- **Only use what a search actually returned.** Every paper title, author, journal, year, and link must come from this session's searches. Anything drawn from training knowledge instead must be labeled `[Not from search — model knowledge]` and excluded from the final list.
- **Confirm before moving on.** A search isn't complete until the results are received and inspected.
- **Track three counts**: queries sent, papers received, papers cited. Surface these in a closing audit summary.
- **Surface gaps, don't fill them.** A section with one paper plus a note about limited results beats a section padded with fabricated entries.

## Phase 0: Grill-Me Intake (3 forcing questions, one per turn)

**Q1 — Syllabus content.** Ask the user to paste the syllabus content (topics, week-by-week breakdown, or learning outcomes) directly into chat. *Why:* there's no file access, so the content has to come through the conversation.

**Q2 — Course audience.** Pick one: undergraduate intro; undergraduate advanced; graduate (Masters/early PhD); graduate (doctoral/advanced); professional/continuing education; mixed. *Why:* this sets summary jargon level and discussion-question complexity — undergrad summaries define every term, grad summaries assume technical fluency; undergrad questions test analysis, grad questions test critique and extension.

**Q3 — Year range.** Pick one: last 1 year (freshest only); last 2 years (default — recent plus a year of context); last 5 years (broader, includes foundational recent work). *Why:* reading lists go stale fast, but a 1-year filter can miss foundational work that's already standard.

Stop at 3 questions before Phase 1. The group-and-confirm checkpoint after Phase 2 is its own separate grill-me moment.

## Phase 1: Parse the Syllabus

From the pasted text, extract: course title, instructor, and term; the topic list (lecture titles or week-by-week breakdown); and learning outcomes. If outcomes aren't explicit, infer 3–5 from the description and mark them `[inferred]` in the final document.

## Phase 2: Group Topics, Then Confirm

Cluster related topics into 6–12 sections (closely-related topics merge; cross-cutting topics get their own section). Then present the proposed sections and ask the user to pick one: looks good, proceed; merge sections X and Y; split section X; add a section for a missing topic; remove a section. *Why this matters:* the grouping drives search allocation — this is the last cheap moment to correct course before search budget is spent. **Do not begin Phase 3 without an explicit user choice.**

## Phase 3: Search Per Section

Search sequentially (roughly one query per second, 1–2 queries per section), constrained by the year range from Q3.

**Applied-domain weaving** — don't just search the bare topic, search topic + applied domain, since this dramatically improves relevance:

| Generic | Applied-domain |
|---|---|
| "enzyme kinetics" | "enzyme kinetics food processing applications" |
| "machine learning" | "machine learning clinical decision support" |
| "thermodynamics" | "thermodynamics renewable energy systems" |
| "social network analysis" | "social network analysis public health interventions" |

Per section: construct the applied-domain query, search, and if results are thin, run one fallback query without the applied-domain angle. Select 1–3 papers per section (15–25 papers total across all sections), prioritizing: relevance to the section topic; reviews/meta-analyses that synthesize the field; citation count/established work; and applied-domain connection to the course.

## Phase 4: Write Summaries + Discussion Questions

**Summaries**: 2–3 sentences, plain language calibrated to the audience from Q2 — define jargon for undergrads, assume fluency for grad audiences. Good: "This review maps how different diets — Mediterranean, Nordic, vegetarian — reshape the types of fat molecules circulating in your blood, with implications for heart disease risk." Bad (too jargon-heavy for most audiences): "This paper reviews lipidomic profiles across dietary interventions and their cardiometabolic implications."

**Discussion questions**: Bloom **higher-order** (apply/analyze/evaluate), tied to a specific course learning outcome, and designed to promote discussion rather than recall. Good: "If dietary fat quality can reshape your lipoprotein lipidome, what does this suggest about the biochemical basis for dietary guidelines recommending unsaturated over saturated fats?" Bad: "What did the authors find?" (pure recall).

## Phase 5: Deliver

Return the reading list directly in chat as formatted markdown: title page info, intro, learning outcomes (with `[inferred]` flags where applicable), then each section with its papers — title, authors, journal, year, link, summary, and discussion question. Close with the audit summary: "{N} sections × {M} papers / {K} cited," plus queries sent / papers received / papers cited.

## Error Handling

| Failure | Behavior |
|---|---|
| A search returns 0 results for a section | Note the section as "limited results — consider manual supplementation" |
| 3 consecutive search failures | Stop, alert the user, share what's collected so far |
| Syllabus format the user pastes is unusable (e.g. garbled OCR) | Ask them to paste cleaner text or a topic list |
| Learning outcomes can't be extracted | Infer 3–5 from the course description; mark `[inferred]` |

## Anti-Patterns To Reject

- Searching topics without the applied-domain angle (weak relevance).
- Padding sections with fabricated entries when search results are thin.
- Generic discussion questions ("What did the authors find?").
- Jargon-heavy summaries unsuited to the course's audience level.
- Skipping the group-and-confirm step (wastes search budget on bad clusters).
- Truncating links in the final list — always give the full URL.