confident-tracing
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides documentation and workflow instructions for the confident-trace SDK, which is a tool developed by the author (Confident AI) for AI observability. The behavior described matches the intended purpose of providing span-based tracing for LLM applications.
- [EXTERNAL_DOWNLOADS]: The skill references standard installation procedures for vendor packages (confident-trace via pip and npm) and links to official documentation on GitHub (github.com/confident-ai/confident-trace). These references are consistent with the vendor's identity and follow expected software development practices.
- [INDIRECT_PROMPT_INJECTION]: The skill instruments data from AI models and retrieval systems. Ingestion points include LLM inputs, outputs, and retrieval contexts as documented in references/tracing.md. The skill lacks dangerous capabilities such as file system writing or command execution, limiting the impact of malicious data ingestion. Sanitization is addressed through explicit instructions to avoid tracing secrets and the provision of content capture controls (e.g., capture_content=False).
Audit Metadata