The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a system that processes data from external notes and their properties, which serves as a vector for indirect prompt injection.
Ingestion points: Data is read from note frontmatter properties (e.g., status, author) and file metadata (e.g., file.name, file.path, file.tags) as described in the filters and formulas sections of SKILL.md.
Boundary markers: The skill does not instruct the agent to use specific delimiters or boundary markers when interpolating note data into formulas, nor does it provide warnings to ignore embedded instructions.
Capability inventory: The formula language supports logic (if), string manipulation, and rendering through functions like html() and link(). It also allows fetching file objects via file(path).
Sanitization: While an escapeHTML() function is included in the references/FUNCTIONS_REFERENCE.md, its use is not explicitly required or demonstrated for untrusted data in the main workflow.
[DYNAMIC_EXECUTION]: The skill documents a domain-specific formula language that allows for the runtime evaluation of expressions and rendering of dynamic content.
Evidence: The formulas section in .base files, described in SKILL.md, executes expressions at runtime to compute values from note properties for display.
Evidence: The html() function defined in references/FUNCTIONS_REFERENCE.md allows for the direct rendering of string content as HTML within the Obsidian application view.

obsidian-bases