multi-agent-orchestration

Warn

Audited by Socket on Aug 21, 2026

2 alerts found:

SecurityAnomaly
SecurityMEDIUM
SKILL.md

该技能的能力与“多 agent 本地编排”目的基本一致,未见明显恶意窃密或隐蔽外传。但它赋予 agent 较强的 shell/git/tmux/多后端 CLI 执行能力,并允许自动提交、PR、合并与凭证环境转发;对 AI skill 而言属高影响编排工具,应视为可疑偏高风险但非恶意。

Confidence: 85%Severity: 72%
AnomalyLOW
scripts/terminal-split.sh

No clear evidence of supply-chain malware (no exfiltration, persistence, credential theft, or concealed payload behavior) is present in this fragment. However, this module is a high-impact automation primitive: it can execute arbitrary, caller-provided COMMAND inside other terminal/editor applications via clipboard and UI/keystroke injection, and it interpolates COMMAND into AppleScript strings without visible escaping. Treat COMMAND as highly sensitive input and only use trusted callers; review/patch for AppleScript string escaping/quoting robustness if distributing broadly.

Confidence: 62%Severity: 62%
Audit Metadata
Analyzed At
Aug 21, 2026, 04:26 PM
Package URL
pkg:socket/skills-sh/cat-xierluo%2Flegal-skills%2Fmulti-agent-orchestration%2F@a84a759b989515fd34fff1776517a8af9dca51d72febd104388edf334a8c9022
Security Audit — socket — multi-agent-orchestration