The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process data from external sources, including Zarr arrays, HDF5 datasets, and metadata from scientific repositories like ArXiv. This represents a potential surface for indirect prompt injection if the ingested data contains malicious instructions targeted at the agent. \n
Ingestion points: The skill uses zarr.open, xr.open_zarr, da.from_zarr, and h5py.File to load external data into the agent's context. It also specifically instructs the agent to fetch and process metadata from arxiv.org. \n
Boundary markers: There are no explicit prompt boundary markers or instructions to ignore embedded commands provided for the content of the ingested datasets. \n
Capability inventory: The skill has access to Bash, Write, and Edit tools, allowing it to execute shell commands and modify files based on instructions interpreted from data. \n
Sanitization: No sanitization, escaping, or validation logic is defined for the external content before it is processed by the agent. \n- [COMMAND_EXECUTION]: The skill provides instructions for executing shell commands via the Bash tool for package management using uv pip install and data migration tasks using the zarr migrate CLI. These commands are standard for the development and maintenance workflows associated with the library and are clearly documented. \n- [EXTERNAL_DOWNLOADS]: The skill references external resources including official documentation, GitHub repositories, and scientific papers on arxiv.org for citations. All referenced domains are well-known and legitimate services for research and software development.

zarr-python