The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Python scripts (build_literature_canvas.py and build_literature_graph.py) to process Markdown files and generate Obsidian Canvas and Mermaid diagrams within the user's project vault.
[DYNAMIC_EXECUTION]: The Python scripts use sys.path.insert to dynamically load a local utility module (project_kb) from the obsidian-project-memory skill directory. This is used for vault path resolution and metadata handling.
[INDIRECT_PROMPT_INJECTION]: The skill processes paper notes which may contain content from external sources (e.g., abstracts). While the scripts only use this for visualization, the presence of untrusted data in the agent's context constitutes a potential attack surface if the agent later treats that content as instructions.
[DATA_EXPOSURE]: The skill reads user notes in the Papers/ and Knowledge/ directories to extract link and metadata information for the purpose of visualization. This data remains within the local filesystem.

obsidian-literature-workflow