external-cannbot-ops-triton-latency-optimizer

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill instructions and reference materials are focused on performance optimization of Triton operators for NPU hardware. No malicious instructions, obfuscation, or unauthorized data access patterns were found.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided Triton code through the input parameter code_file_path. While processing external code presents a surface for indirect prompt injection (e.g., instructions hidden in comments), the skill defines strict optimization rules, a specific execution order, and a code quality checklist (references/checklist.md) to maintain control over the agent's output and verify accuracy.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:53 PM
Security Audit — agent-trust-hub — external-cannbot-ops-triton-latency-optimizer