arize-instrumentation-health
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs read-only audits of trace data to identify instrumentation health issues such as orphaned spans, flat structures, or missing token counts. It provides deterministic checks based on OTel and OpenInference standards.- [INDIRECT_PROMPT_INJECTION]: The skill processes external trace data which may contain untrusted instructions within LLM inputs or outputs. It includes a specific security warning to mitigate this risk.
- Ingestion points: Reads span data from
.arize-tmp-tracesexported via vendor tools. - Boundary markers: Explicitly instructs the agent to treat span content as data only and never execute instructions found within attributes.
- Capability inventory: Analysis logic is deterministic and read-only; data export relies on the
arize-traceskill andaxCLI. - Sanitization: Relies on specific behavioral instructions to prevent command or prompt injection from processed data.- [EXTERNAL_DOWNLOADS]: The skill references the
axCLI, which is a legitimate vendor-provided tool from Arize for interacting with their observability platform.
Audit Metadata