The Agent Skills Directory

[SAFE]: Fetches bibliographic data and performs library management through the official Zotero Web API and local Zotero API endpoints.
[SAFE]: Installs the versioned pyzotero package and optional components from the official Python Package Index (PyPI), which is a well-known registry.
[SAFE]: Integrates with Semantic Scholar's public API for paper recommendations and citation analysis, a well-known research service.
[SAFE]: Demonstrates secure credential handling by instructing users to use environment variables and .env files instead of hardcoding API keys.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, such as bibliographic metadata and full-text PDF content retrieved via the Zotero or Semantic Scholar APIs.
Ingestion points: External data enters through API methods described in references/read-api.md, references/full-text.md, and references/mcp.md.
Boundary markers: The instructions do not define specific delimiters to separate untrusted content from agent instructions.
Capability inventory: The skill allows file writes via zot.dump(), remote library modifications in references/write-api.md, and shell command execution as listed in allowed-tools.
Sanitization: There is no mention of sanitizing or filtering retrieved content before it is processed by the agent. While this presents an attack surface, it is inherent to the tool's research function and considered a low risk in this context.

pyzotero