The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and display trace and span data from the Arize platform. This data includes user-generated content such as LLM inputs, outputs, and retrieval documents, which constitutes a surface for indirect prompt injection attacks.
Ingestion points: Data enters the agent's context through ax spans export and ax traces export commands as described in SKILL.md.
Boundary markers: The skill includes a specific 'untrusted content guardrail' in SKILL.md that explicitly instructs the agent to treat exported data as raw text and not to interpret it as instructions.
Capability inventory: The skill utilizes the ax CLI to perform network operations (connecting to api.arize.com and flight.arize.com) and file system operations (writing to the .arize-tmp-traces directory).
Sanitization: The skill relies on the provided instructional guardrail to prevent the agent from acting on malicious content embedded in the logs; no programmatic sanitization of the JSON content is mentioned.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the arize-ax-cli tool. The documentation in references/ax-setup.md provides instructions for installing or upgrading this package via standard registries using uv, pipx, or pip.
[PERSISTENCE]: In references/ax-profiles.md, the skill suggests that users persist their configuration, such as the ARIZE_SPACE identifier, by adding export commands to shell profile files like ~/.zshrc or ~/.bashrc. While this is a common practice for CLI configuration, it involves modifying shell startup scripts.

arize-trace