The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of data from an Obsidian vault, which includes user-controlled notes and frontmatter.
Ingestion points: Data is read from note properties (file.properties), tags (file.tags), and file metadata (file.path, file.size) in SKILL.md.
Boundary markers: The schema does not define strict delimiters or instructions for the agent to ignore potentially malicious content embedded in note properties.
Capability inventory: The agent is empowered to write and edit .base files, which include logic structures like formulas and filters as defined in the Schema section of SKILL.md.
Sanitization: While an escapeHTML() function is provided in the Functions Reference, its use is not mandatory in the provided examples, potentially allowing malicious data in notes to influence the rendering of views (e.g., via the html() function).

obsidian-bases