The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes the uv package manager to install the pyzotero library and its optional extras (CLI and MCP) from the official Python Package Index (PyPI).
[CREDENTIALS_UNSAFE]: Authentication is handled via environment variables or .env files, preventing the hardcoding of Zotero API keys and User IDs within the skill's code.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it retrieves and processes untrusted bibliographic data.
Ingestion points: Untrusted data enters the agent context through bibliographic metadata (titles, creators, abstracts) and PDF full-text content retrieved from Zotero libraries, as described in references/read-api.md and references/full-text.md.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the retrieved bibliographic data as untrusted content or to ignore embedded instructions.
Capability inventory: The skill provides significant operational capabilities, including filesystem write access via zot.dump(), library modification via zot.create_items() and zot.delete_item(), and the ability to execute shell commands via the Bash tool and CLI extras.
Sanitization: The skill does not implement or document any sanitization or validation processes for retrieved bibliographic data before it is processed by the agent.

pyzotero