---
name: reading-list
description: "Use when the user asks what a researcher should be reading, wants a curated bibliography for a researcher's interests, asks 'what should X read', 'find me adjacent work to X's research', or wants to know what literature surrounds an author's body of work. Triggers on reading-list, bibliography, or 'what should I read next' style requests anchored on a specific researcher."
---

# Researcher Reading List

Build a curated reading list for a researcher, organized by intellectual thread, by triangulating off their most representative recent and most-cited work.

## Input

The user provides an author name (e.g., "David W. Hogg", "Yoshua Bengio", "Jennifer Doudna").

## Tool Chain

Use the Valency Bond MCP tools available in the current host. Tool names may be
qualified by an MCP server prefix; call the matching exposed tool. If the
required Valency Bond tools are unavailable, say so and stop.

### Step 1: Verify the author exists

Call `get_author_profile` with:
- `author` (string): the author name provided by the user

If no results are found, tell the user the author was not found and suggest checking the spelling or trying a partial name. Stop here.

Note the author's top categories from this result — you'll use them later to interpret which themes the reading list is covering. Also note the resolved author name (`resolved_name`) — use it consistently in subsequent calls and when filtering self-citations.

### Step 2: Get top-cited papers (career anchors)

Call `search_by_author` with:
- `author` (string): the author name
- `limit` (integer): 5
- `sort_by` (string): "citations"
- `strict_mode` (string): "fuzzy"

These are the author's most influential works — they anchor the "what defined their career so far" threads.

### Step 3: Get recent papers (current direction)

Call `search_by_author` with:
- `author` (string): the author name
- `limit` (integer): 5
- `sort_by` (string): "relevance"
- `strict_mode` (string): "fuzzy"

The default `relevance` sort orders by recency. These represent the author's *current* intellectual direction, which may differ from their most-cited work.

### Step 4: Select 3–5 representative seed papers

From the combined Step 2 + Step 3 results, select 3 to 5 **seed papers** that best represent the author's distinct intellectual threads:

- Always include the single most-cited paper (career anchor).
- Always include the most recent paper that has a substantive abstract (current direction).
- Fill the remaining slots by picking papers whose category lists differ — i.e. cover different threads of the author's work, not five papers in the same subfield.

If two papers have nearly identical category lists and similar topics, pick only one.

### Step 5: Find similar papers for each seed

For each seed paper from Step 4, call `find_similar_papers` with:
- `paper_id` (string): the seed paper's ID
- `limit` (integer): 10
- `include_abstract` (boolean): false

If a seed paper has no embedding and the call returns an error, skip that seed and continue with the others. Note the skip in the output.

### Step 6: Aggregate and clean

After all `find_similar_papers` calls complete:

1. **Tag each result** with the seed paper that surfaced it.
2. **Deduplicate** — if a paper appears in multiple seeds' similar lists, keep the entry with the highest similarity score and merge the seed tags.
3. **Filter out self-citations** — drop any paper whose author list contains the focal author (use the `resolved_name` from Step 1 to match, since name normalization may matter).
4. **Group by seed** — partition the cleaned results by which seed paper(s) surfaced them. This grouping defines the reading-list "threads."

## Output Format

### Researcher Summary

A short block:
- **Name**: full name as it appears in the corpus
- **Total papers**: from Step 1
- **Primary domains**: top 3 categories from Step 1

### Reading List by Thread

For each seed paper from Step 4, produce a thread section:

#### Thread N: <one-line characterization of the thread>

Anchored by the seed paper:
- **Seed**: Title (paper ID), year, categories

Recommended reading (5–8 papers per thread, ordered by similarity score):

1. **Title** (paper ID) — first 3 authors, year, category — one-sentence reason it matters here
2. ...

If a seed paper was skipped because it had no embedding, add a note under that thread: *"This seed had no embedding; no similar papers could be retrieved."*

### Cross-Thread Highlights

If any paper appeared as similar to **multiple** seeds, highlight it as a cross-thread find — these are often the most interesting recommendations because they sit at the intersection of the author's threads:

- **Title** (paper ID) — surfaced by Seed A *and* Seed C — why this matters

If no cross-thread papers were found, omit this section.

### Suggested Follow-ups

- Ask for a profile of `<author>` for any author that appears across multiple recommendations.
- Ask for papers similar to `<paper_id>` to dig deeper into any specific recommendation.
- Ask for fresh collaborators for `<author>` when the user is interested in *who* (not what) to engage with next.