computer-vision-expert

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFENO_CODEMETADATA_POISONING
Full Analysis
  • [SAFE]: The skill consists entirely of informational markdown documentation. It contains no executable scripts, shell commands, network request triggers, or sensitive file access patterns.\n- [NO_CODE]: No functional code, binaries, or automated tool configurations were detected within the skill package, precluding any risk of malicious execution.\n- [METADATA_POISONING]: The skill uses deceptive metadata, claiming a futuristic date of 2026 and referencing non-existent software versions such as YOLO26 and SAM 3. These inaccuracies are stylistic elements of a thematic persona and do not introduce security vulnerabilities or facilitate data exfiltration.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 03:34 AM
Security Audit — agent-trust-hub — computer-vision-expert