The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines a workflow with an indirect prompt injection attack surface. It is designed to read and process the contents of arbitrary Obsidian notes to extract entity-relation triples.
Ingestion points: Reads vault notes provided via the $ARGUMENTS parameter using the Read tool.
Boundary markers: Absent. The skill documentation does not outline the use of delimiters or specific instructions to the model to ignore potential commands embedded within the source notes.
Capability inventory: The skill has access to the Bash and Write tools, which are used to modify note frontmatter and manage local directory structures.
Sanitization: While the skill mentions validating the model's JSON output against a schema, it lacks descriptions of sanitization or filtering for the extracted text content.
[DATA_EXPOSURE]: The skill's implementation example contains a hardcoded absolute file path (/Users/ethanphan/Documents/my-obsidian-v1). This discloses information regarding the directory structure and username of the development environment.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform administrative tasks such as creating storage directories (.graph/nodes, .graph/edges) and is designed to execute a helper script (extract_graph.py) to process data.

extract-graph