The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions specify using the bbedit shell command to automatically open generated bibliography files. This allows the agent to execute local applications on the host system.\n- [DATA_EXFILTRATION]: The skill hardcodes an absolute local file path (/Users/niyaro/Desktop/) for saving output. This references a specific local username and involves file system access to the user's Desktop directory.\n- [PROMPT_INJECTION]: The skill processes untrusted metadata, notes, and annotations from a Zotero database, creating an indirect prompt injection surface where malicious data could influence agent behavior.\n
Ingestion points: Bibliographic metadata, notes, and annotations retrieved via the zotero_get_item_metadata, zotero_get_annotations, and zotero_search_notes tools (SKILL.md).\n
Boundary markers: Absent; data from Zotero is directly interpolated into the agent's context and output without delimiters.\n
Capability inventory: Includes local file writing and shell command execution (bbedit).\n
Sanitization: No sanitization or validation of the content retrieved from Zotero is described.

zotero-mcp