The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines workflows that ingest untrusted data from external sources, which could contain malicious instructions designed to influence the agent's behavior during note creation or summarization.
Ingestion points: The web_clipper functionality ingests {{selection}} from browser pages, and the research_workflow processes data from gather_sources based on a user-provided {{topic}}.
Boundary markers: The YAML-based note templates do not include clear delimiters or "ignore instructions" markers to separate untrusted external content from the agent's primary instructions.
Capability inventory: The skill utilizes notes-mcp tools, specifically obsidian_create_note, obsidian_search, and obsidian_link, allowing the agent to write to and modify the user's local knowledge base.
Sanitization: There are no explicit sanitization or filtering steps defined to process the ingested web content before it is used to generate or populate notes.

Obsidian Automation