The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, creating an attack surface for indirect prompt injection.
Ingestion points: Data is loaded from external files (CSV, JSON, JSONL, Parquet) and inline JSON payloads through the ax datasets create and ax datasets append commands.
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores potential instructions embedded within the dataset records.
Capability inventory: The skill utilizes the ax CLI tool, which has the capability to perform network operations, file system reads, and file system writes.
Sanitization: No data sanitization or structural validation is specified for the content of the imported records.
[PERSISTENCE]: The reference documentation in references/ax-profiles.md suggests a persistence mechanism by instructing users to add environment variable exports to shell profile files.
Evidence: The skill recommends adding export ARIZE_SPACE="..." to ~/.zshrc or ~/.bashrc (macOS/Linux) or using SetEnvironmentVariable (Windows) to maintain state across terminal sessions.

arize-dataset