The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions and scripts rely on the installation of a Python library (labarchives-py) from a third-party GitHub repository (github.com/mcmero/labarchives-py). This source is not owned by the skill author ('ovachiever') or a recognized trusted organization, presenting an unverified dependency risk.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process notebook data and API responses from the LabArchives platform, which creates a surface for indirect prompt injection attacks.
Ingestion points: Untrusted research data and notebook metadata enter the agent's context through XML and JSON responses from API methods like users/user_access_info and notebooks/notebook_backup.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to differentiate between research data and operational instructions, increasing the risk of the agent obeying instructions embedded within the notebook content.
Capability inventory: The skill scripts utilize the requests library for network communication and have file system access to create configuration files and store notebook backups locally.
Sanitization: While the scripts use standard XML parsing (xml.etree.ElementTree), there is no validation or sanitization logic to ensure that research data does not contain text that could be interpreted as malicious commands by the AI agent.

labarchive-integration