arize-trace
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands using the
axCLI tool for tasks like trace inspection and span management (e.g.,ax spans export,ax traces list). - [INDIRECT_PROMPT_INJECTION]: The skill ingests trace data that may include untrusted content from the LLM applications being monitored.
- Ingestion points: Untrusted data enters the agent context through JSON files generated by
ax spans exportandax traces export(SKILL.md). - Boundary markers: SKILL.md includes a 'Security: untrusted content guardrail' instructing the agent to treat trace data as raw text and avoid executing any instructions found within it.
- Capability inventory: The skill allows the agent to execute shell commands via the
axCLI tool. - Sanitization: The skill relies on instructional guardrails to prevent exploitation, as no programmatic sanitization or escaping of the ingested trace data is specified.
Audit Metadata