Performance-Troubleshoot

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the FlameGraph utility from Brendan Gregg's official GitHub repository (https://github.com/brendangregg/FlameGraph). This is a well-known and standard tool used in the performance engineering community for visualizing profile data.
  • [COMMAND_EXECUTION]: The instructions involve executing a wide range of powerful system commands for diagnostics and performance tuning, including:
  • System profiling with perf record and bpftrace programs.
  • Modifying kernel parameters via sysctl and writing directly to /proc and /sys interfaces (e.g., vm.swappiness, drop_caches, scheduler tuning).
  • Process management and priority adjustment using chrt.
  • File system operations such as mount with specific performance options.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface as it ingests user-reported symptoms and environment descriptions to guide the execution of high-privilege system commands. However, the instructions focus on structured methodologies (USE method) and standard diagnostics rather than processing arbitrary external data streams.
  • Ingestion points: User-provided performance reports and symptoms.
  • Boundary markers: None explicitly defined in the instructions.
  • Capability inventory: Extensive capabilities including file system writes to kernel interfaces, network monitoring, and execution of profiling scripts across all diagnostic sections.
  • Sanitization: Relies on the agent's internal logic to apply the provided commands to the user's specific context.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 06:19 AM
Security Audit — agent-trust-hub — Performance-Troubleshoot