The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could theoretically contain malicious instructions, but it includes appropriate safety guardrails.
Ingestion points: The skill ingests JSON record sources via the ax annotation-queues add-records command and DataFrames through the arize Python SDK update_annotations method.
Boundary markers: The skill includes specific directives instructing the agent to never read .env files, search the filesystem for credentials, or ask the user to paste secrets into the chat interface.
Capability inventory: Capabilities are focused on network operations to the Arize API via the CLI and SDK for annotation management.
Sanitization: Data is validated against project-specific annotation schemas defined within the Arize space.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of official software from the vendor.
Evidence: The skill instructions refer to installing the arize-ax-cli and using the arize Python package. These are verified official tools provided by the vendor (arize-ai) for interacting with their platform.
[COMMAND_EXECUTION]: The skill utilizes command-line tools to interact with remote platform resources.
Evidence: The skill provides examples for using the ax CLI to list, create, and update annotation configurations and queues, which are standard operations for the tool's intended purpose.

arize-annotation