The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users and the agent to download and install a Python package (labarchives-py) from a public GitHub repository (github.com/mcmero/labarchives-py) that is not identified as a trusted source.
[REMOTE_CODE_EXECUTION]: By requiring the installation of an unverified package from a third-party repository at runtime, the skill creates a risk of executing unvetted code in the local environment.
[COMMAND_EXECUTION]: The skill provides several Python scripts (setup_config.py, notebook_operations.py, entry_operations.py) designed to be executed via the shell for configuration, notebook backups, and data management tasks.
[INDIRECT_PROMPT_INJECTION]: The skill provides the capability to ingest and process notebook content, entry data, and metadata from the LabArchives API, which represents an indirect prompt injection surface.
Ingestion points: LabArchives API responses containing notebook entries, comments, and metadata handled in scripts/notebook_operations.py and scripts/entry_operations.py.
Boundary markers: Absent; there are no instructions for the agent to treat data retrieved from the API as untrusted or to ignore embedded instructions within notebook content.
Capability inventory: Network access (requests.post), file system write access (backups and file creation), and execution of local scripts.
Sanitization: Absent; the skill retrieves, displays, and stores entry content (often HTML) and metadata directly from the API without validation or filtering for malicious instructions.

labarchive-integration