langfuse
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Instructions include using
npx @langfuse/cliandnpx langfuse-clifor interacting with the Langfuse platform, which are standard operations for managing LLM observability infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: The skill manages and processes LLM traces, prompt templates, and datasets, constituting an ingestion surface for potentially untrusted data from external observations.\n - Ingestion points:
TRACING.md(ingestion and inspection of observations),PROMPTS.md(variable compilation in templates),EVALUATIONS.md(dataset case inputs and human annotations).\n - Boundary markers: The skill explicitly specifies the use of double-brace syntax for prompt variables to delimit dynamic content from instruction templates.\n
- Capability inventory: The skill utilizes shell command execution via CLI tools and performs data operations through the Langfuse API (e.g., POST /api/public/v2/evaluators).\n
- Sanitization: Instructions mandate following application-specific telemetry privacy rules and content controls at the point of capture to prevent unauthorized data exposure.
Audit Metadata