dotnet-trace-collect

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • External Downloads: The skill provides instructions to download diagnostic utilities like perfcollect, dotnet-trace, and PerfView from Microsoft-managed domains (aka.ms) and official GitHub repositories. These are standard tools for .NET performance analysis and originate from a trusted organization.
  • Command Execution and Privilege Escalation: To capture kernel-level events and machine-wide traces, the skill utilizes sudo or root permissions for various commands. It also describes setting up privileged sidecar containers in Kubernetes. These elevated privileges are a standard requirement for performing low-level system and runtime diagnostics.
  • Indirect Prompt Injection: The skill ingests user-defined parameters such as process IDs (PIDs) and symptom descriptions, which are then interpolated into shell commands. This represents a potential surface where malformed input could influence command execution.
  • Ingestion points: Inputs defined in SKILL.md such as Symptom, Runtime, and OS, as well as PID values requested during the workflow.
  • Boundary markers: The prompt instructions do not use specific delimiters or protective framing when interpolating user-provided values into command strings.
  • Capability inventory: The skill is designed to generate and potentially execute shell commands (some requiring sudo) across several reference files (perfcollect.md, dotnet-trace-collect-linux.md, etc.).
  • Sanitization: The skill advises manual verification of the target process (e.g., using dotnet-trace ps) before collection, which serves as a manual check for input validity.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 07:06 PM
Security Audit — agent-trust-hub — dotnet-trace-collect