llm-obs-trace-rca

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the pup CLI tool via shell commands in Bash to perform diagnostics, such as searching for spans and retrieving trace structures. This is a primary method for interacting with the backend services.
  • [DATA_EXFILTRATION]: The skill transmits and receives data from official Datadog endpoints (datadoghq.com) to analyze performance and export diagnostic reports. These operations are directed to well-known infrastructure associated with the skill's author and purpose.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and analyzes potentially untrusted content from production traces, including LLM conversation history and retrieved RAG documents.
  • Ingestion points: External trace content is retrieved using the get_llmobs_span_content tool throughout the analysis workflow.
  • Boundary markers: The skill instructions do not define the use of delimiters or specific safety guidelines (e.g., 'ignore embedded instructions') when processing fetched trace data.
  • Capability inventory: The agent has access to shell command execution and local file writing capabilities.
  • Sanitization: No explicit sanitization or validation of the retrieved trace content is mentioned in the skill methodology.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 08:21 PM
Security Audit — agent-trust-hub — llm-obs-trace-rca