claude-agent-sdk
Warn
Audited by Socket on Oct 25, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The skill's purpose is coherent as SDK documentation, and there is no direct exfiltration or malicious execution. However, the installation instructions appear inconsistent with the official package names in the provided evidence, and the MCP examples invoke unpinned npx packages, so users could install the wrong dependencies or forward execution trust to third-party code. This looks more like inaccurate or stale documentation than malware.
Confidence: 91%Severity: 56%
Audit Metadata