Zotero Cli

Command-line interface for Zotero - search your Zotero library, add/edit notes, read attachments, and manage bibliographic references from the terminal.

This skill is internally consistent with a Zotero CLI: it uses Python and installs via pip/pipx, and includes helper scripts for searching, exporting, backup, and update checks. Before installing or running: 1) verify the upstream repository (the SKILL lists a GitHub homepage) and that the package on PyPI/GitHub matches what you expect; 2) inspect the scripts (setup_and_check.sh, backup_restore.sh, update_check.sh) because they can run installation and networking operations and backup your config — backup_restore.sh explicitly stores ~/.config/zotcli/config.ini which contains your Zotero userID/API key, so ensure backups are written to a secure location with restrictive permissions; 3) prefer pipx or a virtualenv rather than system pip, and avoid dangerous flags like --break-system-packages; 4) if you plan to automate (cron) or run update checks, review those scripts to confirm what network endpoints they contact (PyPI/GitHub are expected) and whether they perform upgrades automatically. If you want extra assurance, review the actual script contents (they're included) or run them in a disposable VM/container first.

Type: OpenClaw Skill Name: zotero-cli Version: 0.1.0 The OpenClaw AgentSkills skill bundle for 'zotero-cli' is classified as benign. All scripts and documentation are clearly aligned with the stated purpose of managing Zotero references via a command-line interface. Risky capabilities such as `sudo` are used for legitimate system package installations (e.g., `pipx`, `pandoc`) within setup and update scripts (`setup_and_check.sh`, `update_check.sh`). Network calls are limited to fetching version information from PyPI and GitHub, and an example of `scp` for sharing bibliographies, all of which are transparent and serve the tool's intended functionality. There is no evidence of intentional malicious behavior, data exfiltration to unauthorized endpoints, or prompt injection attempts against the agent to perform harmful actions.