confident-otel
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instruments AI applications to capture and export LLM inputs, outputs, and tool calls. This data ingestion path creates a surface for indirect prompt injection where malicious content in the processed traces could target downstream analysis systems. As an observability tool, this behavior is central to its functionality.\n
- Ingestion points: Captured traces from agent loops, LLM calls, and retrieval steps (e.g., confident.span.input in SKILL.md, confident.trace.output in templates/confident_otel_setup.py).\n
- Boundary markers: Not specified; traces are captured raw from the AI runtime.\n
- Capability inventory: Network egress via OTLP/HTTP exporters to vendor endpoints.\n
- Sanitization: Instructions recommend JSON-encoding non-primitive data types but do not include content sanitization filters for the trace payloads.\n- [EXTERNAL_DOWNLOADS]: The skill recommends installing official, industry-standard OpenTelemetry libraries (opentelemetry-sdk and opentelemetry-exporter-otlp-proto-http) from public registries to enable tracing capabilities.\n- [DATA_EXFILTRATION]: The skill facilitates the transmission of application telemetry data to the vendor's cloud infrastructure (otel.confident-ai.com). The instructions correctly emphasize using environment variables for API keys and utilizing HTTPS for secure data transport.
Audit Metadata