The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches academic paper metadata, abstracts, and citation counts from DBLP (dblp.org) and Semantic Scholar (api.semanticscholar.org) to build recommendations.
[COMMAND_EXECUTION]: Executes several local Python scripts to scan existing notes, search for new papers, and link keywords within the user's Obsidian vault environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the processing of untrusted paper titles and abstracts retrieved from external academic APIs.
Ingestion points: External data sources including DBLP and Semantic Scholar API responses containing paper titles, abstracts, and author names.
Boundary markers: There are no explicit delimiters or instructions to treat retrieved metadata as untrusted data or to ignore embedded instructions within the metadata.
Capability inventory: The skill has the capability to write files to the local file system (Obsidian vault) and trigger subsequent analysis skills such as paper-analyze and extract-paper-images based on the retrieved content.
Sanitization: Content retrieved from the academic APIs is used directly in markdown file generation without evidence of sanitization or validation to prevent instruction injection.

conf-papers