vault-graph

Name: vault-graph
Author: tuan3w

32stars2forks

How to install vault-graph?

tuan3w/obsidian-vault-agent/vault-graph

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

Installs into the current project.

Summary

Use when analyzing vault structure, finding orphan notes, discovering missing connections, identifying bridge concepts, or checking vault health from a graph perspective. Triggers on "vault graph", "map vault", "find orphans", "missing links", "vault structure", "knowledge graph"

Files · 1

View on GitHub

SKILL.md

1	<Purpose>
2	Analyze the vault's wikilink structure as a directed graph. Computes PageRank
3	(most influential concepts), betweenness centrality (bridge nodes connecting
4	domains), orphan detection, cluster analysis, and missing-link discovery.
5	Combines deterministic graph computation (Python/NetworkX) with intelligent
6	interpretation (agent) to surface actionable insights.
7	</Purpose>
8
9	<Use_When>
10	- User asks about vault structure, connections, or health
11	- User wants to find orphan notes (unlinked, isolated)
12	- User wants to discover missing connections between notes
13	- User wants to identify the most important concepts in the vault
14	- User asks "what should I link?" or "what's disconnected?"
15	- As part of /health workflow for structural metrics
16	- User says "map", "graph", "network", "connections", "orphans", "bridges"
17	</Use_When>
18
19	<Do_Not_Use_When>
20	- User wants to search note CONTENT (use Grep/search_notes instead)
21	- User wants to process a single note (use /process)
22	- User wants tag analysis without graph structure (use /health)
23	</Do_Not_Use_When>
24
25	<Execution_Policy>
26	- Run the Python script first — it handles all graph math deterministically
27	- Pipe script output to the graph-analyst agent for interpretation
28	- Report progress via TodoWrite
29	- If uv is not installed, provide install command and stop
30	- Script runs on the full vault — no sampling needed (handles 5000+ notes)
31	</Execution_Policy>
32
33	<Steps>
34
35	## Stage 1: RUN GRAPH ANALYSIS
36
37	Run the analysis script from the skill directory:
38
39	```bash
40	SKILL_DIR="${CLAUDE_SKILL_DIR}"
41	uv run "$SKILL_DIR/scripts/analyze_vault_graph.py" "." --top 20
42	```
43
44	The script outputs JSON to stdout with:
45	- `summary`: total notes, edges, orphans, density, clustering coefficient
46	- `top_pagerank`: most influential notes
47	- `top_betweenness`: bridge concepts connecting clusters
48	- `top_in_degree`: most referenced notes
49	- `top_out_degree`: most connecting notes
50	- `orphans`: notes with zero links in/out
51	- `dead_ends`: notes referenced but linking nowhere
52	- `clusters`: connected components with dominant types
53	- `missing_links`: wikilinks pointing to non-existent notes
54	- `type_distribution`: counts by note type
55	- `status_distribution`: counts by processing status
56
57	Requires `uv` installed (`brew install uv` or `curl -LsSf https://astral.sh/uv/install.sh \| sh`). Dependencies auto-install on first run.
58
59	## Stage 2: INTERPRET WITH AGENT
60
61	Read the agent definition from `agents/graph-analyst.md` in the skill directory.
62
63	Launch the analyst agent:
64
65	```
66	Agent(
67	subagent_type="general-purpose",
68	model="sonnet",
69	run_in_background=false,
70	prompt="You are Graph Analyst. Follow these instructions exactly:
71
72	[INSERT FULL CONTENT OF agents/graph-analyst.md HERE]
73
74	VAULT CONTEXT:
75	- This is a Zettelkasten-style Obsidian vault
76	- Note types: term (atomic concepts), thought (original synthesis),
77	paper/post/book (sources), note (explanations), decision-log
78	- Tags are inline (#topic), not frontmatter
79	- Cross-domain connections are the vault's highest-value links
80
81	GRAPH ANALYSIS RESULTS:
82	[INSERT JSON OUTPUT FROM STAGE 1 HERE]
83
84	Produce your analysis following the Output Format specified above."
85	)
86	```
87
88	## Stage 3: PRESENT RESULTS
89
90	Present the agent's analysis to the user. Include:
91	1. The health score and top findings
92	2. Specific bridge concepts and orphans
93	3. Missing notes worth creating
94	4. Suggested new connections
95
96	If the user wants to ACT on suggestions (create notes, add links), help them
97	do so using standard vault tools (Edit, write_note, etc.).
98
99	</Steps>
100
101	<Tool_Usage>
102	- Bash: Run analyze_vault_graph.py script
103	- Read: Read agent definition from agents/graph-analyst.md
104	- Agent: Delegate interpretation to graph-analyst (sonnet)
105	- TodoWrite: Report progress at each stage
106	- Grep/Glob: Follow-up searches if user wants to explore specific findings
107	</Tool_Usage>
108
109	<Examples>
110	<Good>
111	User: "Show me my vault's knowledge graph health"
112	1. Run script → JSON with 847 notes, 2341 edges, 43 orphans
113	2. Agent interprets → "Your ML cluster is dense but isolated from psychology.
114	[[Reinforcement Learning]] and [[Operant Conditioning]] both describe
115	reward-based behavior but aren't linked."
116	3. Present: 5 findings, 8 bridge suggestions, 12 orphans worth connecting
117	</Good>
118
119	<Good>
120	User: "Find orphan notes"
121	1. Run script → 43 orphans identified
122	2. Agent filters → 15 are legitimately standalone (daily notes, clippings),
123	28 should be connected
124	3. Present prioritized list: "[[Loss Aversion]] is a term note with zero
125	links — should connect to [[Prospect Theory]] and [[Pricing Psychology]]"
126	</Good>
127
128	<Bad>
129	User: "Map m

People who liked this also viewed

6 picks

Type

vault-graph

32stars2forks

View on GitHub

How to install vault-graph?

tuan3w/obsidian-vault-agent/vault-graph

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<Purpose>
2	Analyze the vault's wikilink structure as a directed graph. Computes PageRank
3	(most influential concepts), betweenness centrality (bridge nodes connecting
4	domains), orphan detection, cluster analysis, and missing-link discovery.
5	Combines deterministic graph computation (Python/NetworkX) with intelligent
6	interpretation (agent) to surface actionable insights.
7	</Purpose>
8
9	<Use_When>
10	- User asks about vault structure, connections, or health
11	- User wants to find orphan notes (unlinked, isolated)
12	- User wants to discover missing connections between notes
13	- User wants to identify the most important concepts in the vault
14	- User asks "what should I link?" or "what's disconnected?"
15	- As part of /health workflow for structural metrics
16	- User says "map", "graph", "network", "connections", "orphans", "bridges"
17	</Use_When>
18
19	<Do_Not_Use_When>
20	- User wants to search note CONTENT (use Grep/search_notes instead)
21	- User wants to process a single note (use /process)
22	- User wants tag analysis without graph structure (use /health)
23	</Do_Not_Use_When>
24
25	<Execution_Policy>
26	- Run the Python script first — it handles all graph math deterministically
27	- Pipe script output to the graph-analyst agent for interpretation
28	- Report progress via TodoWrite
29	- If uv is not installed, provide install command and stop
30	- Script runs on the full vault — no sampling needed (handles 5000+ notes)
31	</Execution_Policy>
32
33	<Steps>
34
35	## Stage 1: RUN GRAPH ANALYSIS
36
37	Run the analysis script from the skill directory:
38
39	```bash
40	SKILL_DIR="${CLAUDE_SKILL_DIR}"
41	uv run "$SKILL_DIR/scripts/analyze_vault_graph.py" "." --top 20
42	```
43
44	The script outputs JSON to stdout with:
45	- `summary`: total notes, edges, orphans, density, clustering coefficient
46	- `top_pagerank`: most influential notes
47	- `top_betweenness`: bridge concepts connecting clusters
48	- `top_in_degree`: most referenced notes
49	- `top_out_degree`: most connecting notes
50	- `orphans`: notes with zero links in/out
51	- `dead_ends`: notes referenced but linking nowhere
52	- `clusters`: connected components with dominant types
53	- `missing_links`: wikilinks pointing to non-existent notes
54	- `type_distribution`: counts by note type
55	- `status_distribution`: counts by processing status
56
57	Requires `uv` installed (`brew install uv` or `curl -LsSf https://astral.sh/uv/install.sh \| sh`). Dependencies auto-install on first run.
58
59	## Stage 2: INTERPRET WITH AGENT
60
61	Read the agent definition from `agents/graph-analyst.md` in the skill directory.
62
63	Launch the analyst agent:
64
65	```
66	Agent(
67	subagent_type="general-purpose",
68	model="sonnet",
69	run_in_background=false,
70	prompt="You are Graph Analyst. Follow these instructions exactly:
71
72	[INSERT FULL CONTENT OF agents/graph-analyst.md HERE]
73
74	VAULT CONTEXT:
75	- This is a Zettelkasten-style Obsidian vault
76	- Note types: term (atomic concepts), thought (original synthesis),
77	paper/post/book (sources), note (explanations), decision-log
78	- Tags are inline (#topic), not frontmatter
79	- Cross-domain connections are the vault's highest-value links
80
81	GRAPH ANALYSIS RESULTS:
82	[INSERT JSON OUTPUT FROM STAGE 1 HERE]
83
84	Produce your analysis following the Output Format specified above."
85	)
86	```
87
88	## Stage 3: PRESENT RESULTS
89
90	Present the agent's analysis to the user. Include:
91	1. The health score and top findings
92	2. Specific bridge concepts and orphans
93	3. Missing notes worth creating
94	4. Suggested new connections
95
96	If the user wants to ACT on suggestions (create notes, add links), help them
97	do so using standard vault tools (Edit, write_note, etc.).
98
99	</Steps>
100
101	<Tool_Usage>
102	- Bash: Run analyze_vault_graph.py script
103	- Read: Read agent definition from agents/graph-analyst.md
104	- Agent: Delegate interpretation to graph-analyst (sonnet)
105	- TodoWrite: Report progress at each stage
106	- Grep/Glob: Follow-up searches if user wants to explore specific findings
107	</Tool_Usage>
108
109	<Examples>
110	<Good>
111	User: "Show me my vault's knowledge graph health"
112	1. Run script → JSON with 847 notes, 2341 edges, 43 orphans
113	2. Agent interprets → "Your ML cluster is dense but isolated from psychology.
114	[[Reinforcement Learning]] and [[Operant Conditioning]] both describe
115	reward-based behavior but aren't linked."
116	3. Present: 5 findings, 8 bridge suggestions, 12 orphans worth connecting
117	</Good>
118
119	<Good>
120	User: "Find orphan notes"
121	1. Run script → 43 orphans identified
122	2. Agent filters → 15 are legitimately standalone (daily notes, clippings),
123	28 should be connected
124	3. Present prioritized list: "[[Loss Aversion]] is a term note with zero
125	links — should connect to [[Prospect Theory]] and [[Pricing Psychology]]"
126	</Good>
127
128	<Bad>
129	User: "Map m

People who liked this also viewed

6 picks

Type