Thesis
Adding a second independent scanner can broaden the evidence available to a platform’s security review, but benchmark results should not be read as a guarantee for live uploads.
Observed facts
OpenClaw says every skill and plugin uploaded to ClawHub now runs through Tencent’s AI-Infra-Guard (AIG) as part of review.
1ClawScan runs AIG and NVIDIA’s SkillSpector independently, then has an AI judge consider both findings alongside the uploaded files.
1OpenClaw reports 86.9% agreement with benchmark labels and correct classification of 98.6% of malicious cases on a fixed 556-case subset of SkillTrustBench.
1
Mechanism
The integration preserves distinct scanner assessments for a later review step rather than treating one scanner as the sole verdict.
Why now
AIG is now included in ClawHub’s review path, and the announcement describes a feedback loop in which anonymized disagreements and confirmed false positives are shared with Tencent for regression tests and scanner improvements.
Implications
The integration makes cross-scanner disagreement an input to review and iteration, rather than hiding it behind a single scanner score.
Evidence boundary
These are figures reported by OpenClaw for a specific benchmark subset; they do not show the rate of missed threats or false positives across all live ClawHub uploads.
Watch conditions
Watch for later reporting that measures false positives, missed threats, or changes in benchmark results after the feedback loop updates the scanners.