paper-discover

Name: paper-discover
Author: tuan3w

32stars2forks

How to install paper-discover?

tuan3w/obsidian-vault-agent/paper-discover

$npx -y skills add tuan3w/obsidian-vault-agent --skill paper-discover

Installs into the current project.

Summary

Use when searching for academic papers related to a topic, finding papers similar to one already in the vault, or discovering research gaps. Triggers on "find papers", "related papers", "paper search", "literature", "what papers should I read about X".

Files · 1

View on GitHub

SKILL.md

1	<Purpose>
2	Search academic databases (Semantic Scholar) for papers relevant to a topic
3	or related to an existing vault note. Returns ranked results with relevance
4	assessment against vault content. Optionally creates formatted paper notes
5	in the vault's inbox for processing.
6	</Purpose>
7
8	<Use_When>
9	- User asks to find papers on a topic ("find papers about scaling laws")
10	- User wants papers related to an existing note ("what papers connect to this?")
11	- User is /processing a paper and wants to find related work
12	- User asks "what should I read about X?"
13	- User provides a DOI or paper title and wants similar papers
14	- User wants to fill gaps in a knowledge domain
15	</Use_When>
16
17	<Do_Not_Use_When>
18	- User has a paper file to analyze (use /paper or /book-analyzer)
19	- User wants to process an existing vault note (use /process)
20	- User wants general web research, not academic papers (use /research)
21	</Do_Not_Use_When>
22
23	<Execution_Policy>
24	- Search first, present results, then create notes only if user approves
25	- Always check vault for existing paper notes before creating duplicates
26	- Rank by vault relevance, not just citation count
27	- Cap at 10 results per search — quality over quantity
28	- Respect Semantic Scholar rate limits (100 req/5min)
29	</Execution_Policy>
30
31	<Steps>
32
33	## Stage 1: PARSE QUERY AND CONTEXT
34
35	Determine the search mode from user input:
36
37	Mode A — Topic search (default):
38	User provides a topic or question. Extract search terms.
39	```
40	"find papers about scaling laws for LLMs" → query: "scaling laws large language models"
41	```
42
43	Mode B — Similar papers:
44	User references an existing vault note or provides a paper ID/DOI.
45	1. Read the referenced note to extract title, key concepts
46	2. Use the paper's Semantic Scholar ID or DOI for recommendations
47	3. Fall back to keyword search if no ID available
48
49	Mode C — Gap filling:
50	User asks about a domain. Search vault first to identify what's covered,
51	then search for papers on uncovered subtopics.
52
53	## Stage 2: SEARCH
54
55	Run the search script:
56
57	```bash
58	SKILL_DIR="${CLAUDE_SKILL_DIR}"
59
60	# Topic search
61	python3 "$SKILL_DIR/scripts/search_papers.py" "QUERY" --limit 20 --min-citations 5
62
63	# With year filter
64	python3 "$SKILL_DIR/scripts/search_papers.py" "QUERY" --limit 20 --year-from 2020
65
66	# Get recommendations from a paper
67	python3 "$SKILL_DIR/scripts/search_papers.py" "" --recommend-from "DOI:10.xxxx/xxxxx"
68
69	# Look up specific paper
70	python3 "$SKILL_DIR/scripts/search_papers.py" "" --paper-id "DOI:10.xxxx/xxxxx"
71	```
72
73	Script returns JSON array of papers with: title, authors, year, abstract,
74	tldr, citation_count, influential_citations, url, doi, arxiv_id, fields_of_study.
75
76	## Stage 3: RANK AND PRESENT
77
78	Read the agent definition from `agents/paper-writer.md` in the skill directory.
79
80	Search the vault for existing paper notes on this topic using the MCP tool:
81	```
82	search_notes(query="KEYWORD", limit=20)
83	```
84	Or fall back to Grep if MCP is unavailable:
85	```
86	Grep(pattern="KEYWORD", path="notes/", glob="*.md", head_limit=20)
87	```
88
89	Launch the paper-writer agent to rank and present:
90
91	```
92	Agent(
93	subagent_type="general-purpose",
94	model="sonnet",
95	run_in_background=false,
96	prompt="You are Paper Writer. Follow these instructions exactly:
97
98	[INSERT FULL CONTENT OF agents/paper-writer.md HERE]
99
100	SEARCH CONTEXT:
101	- User query: [original query]
102	- Search mode: [topic/similar/gap]
103	- Vault topics: [detected from search/tags]
104
105	EXISTING VAULT PAPERS:
106	[List any matching paper notes already in the vault]
107
108	SEARCH RESULTS:
109	[INSERT JSON FROM STAGE 2 HERE]
110
111	Rank these papers by relevance to the vault's existing knowledge.
112	Present the top 10 with your assessment. Do NOT create notes yet —
113	just present the ranked list for user triage."
114	)
115	```
116
117	## Stage 4: CREATE NOTES (on user approval)
118
119	After user selects which papers to add:
120
121	For each selected paper, create a vault note:
122	1. Generate timestamp ID: `date +%Y%m%d%H%M%S`
123	2. Create the note file with proper frontmatter and body
124	3. Check for duplicate titles before creating
125
126	Note template:
127	```markdown
128	---
129	id: YYYYMMDDHHMMSS
130	created_date: YYYY-MM-DD
131	updated_date: YYYY-MM-DD
132	type: paper
133	category:
134	link: [paper URL]
135	processing_status: inbox
136	---
137
138	# Title
139	- 🏷️Tags : #paper #topic-tag #MM-YYYY
140
141	[ ](#anki-card)
142
143	## Abstract
144	- [Synthesized abstract bullets]
145
146	## Notes
147
148	## Questions
149
150	## Related links
151	- [Paper URL](url)
152	- [[(Type) Related Vault Note]]
153	```
154
155	Place the note in `notes/paper/` (or appropriate topic folder).
156
157	</Steps>
158
159	<Tool_Usage>
160	- Bash: Run search_papers.py script, generate timestamps
161	- Read: Read agent definition, read existing vault notes for context
162	- Write/Edit:

People who liked this also viewed

6 picks

Type

paper-discover

32stars2forks

View on GitHub

How to install paper-discover?

tuan3w/obsidian-vault-agent/paper-discover

$npx -y skills add tuan3w/obsidian-vault-agent --skill paper-discover

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<Purpose>
2	Search academic databases (Semantic Scholar) for papers relevant to a topic
3	or related to an existing vault note. Returns ranked results with relevance
4	assessment against vault content. Optionally creates formatted paper notes
5	in the vault's inbox for processing.
6	</Purpose>
7
8	<Use_When>
9	- User asks to find papers on a topic ("find papers about scaling laws")
10	- User wants papers related to an existing note ("what papers connect to this?")
11	- User is /processing a paper and wants to find related work
12	- User asks "what should I read about X?"
13	- User provides a DOI or paper title and wants similar papers
14	- User wants to fill gaps in a knowledge domain
15	</Use_When>
16
17	<Do_Not_Use_When>
18	- User has a paper file to analyze (use /paper or /book-analyzer)
19	- User wants to process an existing vault note (use /process)
20	- User wants general web research, not academic papers (use /research)
21	</Do_Not_Use_When>
22
23	<Execution_Policy>
24	- Search first, present results, then create notes only if user approves
25	- Always check vault for existing paper notes before creating duplicates
26	- Rank by vault relevance, not just citation count
27	- Cap at 10 results per search — quality over quantity
28	- Respect Semantic Scholar rate limits (100 req/5min)
29	</Execution_Policy>
30
31	<Steps>
32
33	## Stage 1: PARSE QUERY AND CONTEXT
34
35	Determine the search mode from user input:
36
37	Mode A — Topic search (default):
38	User provides a topic or question. Extract search terms.
39	```
40	"find papers about scaling laws for LLMs" → query: "scaling laws large language models"
41	```
42
43	Mode B — Similar papers:
44	User references an existing vault note or provides a paper ID/DOI.
45	1. Read the referenced note to extract title, key concepts
46	2. Use the paper's Semantic Scholar ID or DOI for recommendations
47	3. Fall back to keyword search if no ID available
48
49	Mode C — Gap filling:
50	User asks about a domain. Search vault first to identify what's covered,
51	then search for papers on uncovered subtopics.
52
53	## Stage 2: SEARCH
54
55	Run the search script:
56
57	```bash
58	SKILL_DIR="${CLAUDE_SKILL_DIR}"
59
60	# Topic search
61	python3 "$SKILL_DIR/scripts/search_papers.py" "QUERY" --limit 20 --min-citations 5
62
63	# With year filter
64	python3 "$SKILL_DIR/scripts/search_papers.py" "QUERY" --limit 20 --year-from 2020
65
66	# Get recommendations from a paper
67	python3 "$SKILL_DIR/scripts/search_papers.py" "" --recommend-from "DOI:10.xxxx/xxxxx"
68
69	# Look up specific paper
70	python3 "$SKILL_DIR/scripts/search_papers.py" "" --paper-id "DOI:10.xxxx/xxxxx"
71	```
72
73	Script returns JSON array of papers with: title, authors, year, abstract,
74	tldr, citation_count, influential_citations, url, doi, arxiv_id, fields_of_study.
75
76	## Stage 3: RANK AND PRESENT
77
78	Read the agent definition from `agents/paper-writer.md` in the skill directory.
79
80	Search the vault for existing paper notes on this topic using the MCP tool:
81	```
82	search_notes(query="KEYWORD", limit=20)
83	```
84	Or fall back to Grep if MCP is unavailable:
85	```
86	Grep(pattern="KEYWORD", path="notes/", glob="*.md", head_limit=20)
87	```
88
89	Launch the paper-writer agent to rank and present:
90
91	```
92	Agent(
93	subagent_type="general-purpose",
94	model="sonnet",
95	run_in_background=false,
96	prompt="You are Paper Writer. Follow these instructions exactly:
97
98	[INSERT FULL CONTENT OF agents/paper-writer.md HERE]
99
100	SEARCH CONTEXT:
101	- User query: [original query]
102	- Search mode: [topic/similar/gap]
103	- Vault topics: [detected from search/tags]
104
105	EXISTING VAULT PAPERS:
106	[List any matching paper notes already in the vault]
107
108	SEARCH RESULTS:
109	[INSERT JSON FROM STAGE 2 HERE]
110
111	Rank these papers by relevance to the vault's existing knowledge.
112	Present the top 10 with your assessment. Do NOT create notes yet —
113	just present the ranked list for user triage."
114	)
115	```
116
117	## Stage 4: CREATE NOTES (on user approval)
118
119	After user selects which papers to add:
120
121	For each selected paper, create a vault note:
122	1. Generate timestamp ID: `date +%Y%m%d%H%M%S`
123	2. Create the note file with proper frontmatter and body
124	3. Check for duplicate titles before creating
125
126	Note template:
127	```markdown
128	---
129	id: YYYYMMDDHHMMSS
130	created_date: YYYY-MM-DD
131	updated_date: YYYY-MM-DD
132	type: paper
133	category:
134	link: [paper URL]
135	processing_status: inbox
136	---
137
138	# Title
139	- 🏷️Tags : #paper #topic-tag #MM-YYYY
140
141	[ ](#anki-card)
142
143	## Abstract
144	- [Synthesized abstract bullets]
145
146	## Notes
147
148	## Questions
149
150	## Related links
151	- [Paper URL](url)
152	- [[(Type) Related Vault Note]]
153	```
154
155	Place the note in `notes/paper/` (or appropriate topic folder).
156
157	</Steps>
158
159	<Tool_Usage>
160	- Bash: Run search_papers.py script, generate timestamps
161	- Read: Read agent definition, read existing vault notes for context
162	- Write/Edit:

People who liked this also viewed

6 picks

Type