The Agent Skills Directory

[SAFE]: All reviewed files contain benign documentation, installation guides, and standard Python/bash snippets for processing research papers and managing Zotero plugins. There are no signs of credential exposure, malicious remote code execution, or prompt injections.
[INDIRECT_PROMPT_INJECTION]: The skill describes pipelines and workflows for extracting text from external academic documents, PDFs, and web URLs. This outlines a data ingestion surface where processing untrusted text documents could potentially introduce adversarial content into an active LLM session if boundaries are not strictly managed.

document-skills