The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from the Zotero Web API, such as item titles, abstracts, and full-text PDF content, which could serve as a vector for indirect prompt injection if the user manages shared libraries or imports malicious references.
Ingestion points: The zot.items(), zot.top(), and zot.fulltext_item() methods in references/read-api.md and references/full-text.md ingest arbitrary string data from Zotero servers.
Boundary markers: The provided code examples do not demonstrate the use of delimiters or 'ignore instructions' markers when processing metadata.
Capability inventory: The skill has access to the Bash tool and file system operations (Read, Write, Edit), allowing it to write citation files or PDF attachments to disk.
Sanitization: No explicit sanitization or validation of metadata fields is shown before they are used in scripts or likely interpolated into agent prompts.
[EXTERNAL_DOWNLOADS]: The skill utilizes the pyzotero library, an established and well-known Python wrapper for the Zotero API. Installation instructions in SKILL.md and references/cli.md use standard package management via uv.

alterlab-pyzotero