AI Agent Skill Threat Landscape
| Date: 2026-04-03 |
Audited by ClawSec |
Analysis of 20792 threats detected across 2103 audited skills.
Threats by Severity
| Severity |
Count |
% |
| CRITICAL |
1051 |
5% |
| HIGH |
5820 |
28% |
| MEDIUM |
4173 |
20% |
| LOW |
9748 |
47% |
Top 15 Threat Types
| Threat |
Count |
Severity |
| LLM Semantic Detection |
15186 |
LOW |
| Dynamic Code Evaluation |
2016 |
HIGH |
| Startup Failure (non-executable) |
1945 |
LOW |
| Outbound Data Transfer |
764 |
HIGH |
| Hidden Command Execution |
342 |
MEDIUM |
| Private Key Extraction |
107 |
CRITICAL |
| Environment Variable Exfiltration |
107 |
CRITICAL |
| Base64 Encoded Payload |
58 |
HIGH |
| Remote Script Execution |
54 |
CRITICAL |
| Execution Crash |
26 |
MEDIUM |
| Shell RC Modification |
24 |
HIGH |
| Cron Job Installation |
22 |
HIGH |
| Obfuscated Code |
20 |
MEDIUM |
| Systemd Service Installation |
19 |
HIGH |
| Cryptocurrency Wallet Access |
19 |
CRITICAL |
Threats by Verdict
| Verdict |
Threats |
Avg Threats/Skill |
| MALICIOUS |
2065 |
12.0 |
| SUSPICIOUS |
10723 |
10.6 |
| SAFE |
7988 |
8.7 |
ClawSec | ClawSearch