高风险行为识别分析工具
Security checks across malware telemetry and agentic risk
Overview
The skill generally matches its stated video risk-analysis purpose, but its report-listing code can query history without the user's open-id filter, which could expose sensitive health or safety reports.
Review the report-history behavior before installing or using this skill. Only submit videos/images you are allowed to share, confirm the active API endpoint is trusted HTTPS, and avoid using the history-list feature until it is fixed to enforce the current user's open-id filter.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A request to list reports could return more health or safety report records than the user intended, including other users' reports if the backend permits it.
The report-listing method defaults open_id to None and explicitly omits the createBy filter when it is absent, which contradicts the stated goal of listing only the specified user's historical risk-analysis reports.
def get_output_analysis_list(self, pageNum=None, pageSize=None, open_id=None):
...
# open_id 检查 - 如果 open_id 为空/None，不添加过滤条件，返回所有报告
data = {}
if open_id:
data["createBy"] = open_idRequire the current open-id to be passed into every history query, fail closed when it is missing, and rely on server-side authorization as well as client-side filtering.
Videos, images, health-risk results, and location/context may be sent to external services or caregivers as part of alerting.
The skill discloses that alerts may share anomaly footage, risk type, and location information through external notification channels.
通过APP推送、短信、语音播报等多渠道通知家属及护理人员，并同步发送异常画面、风险类型与位置信息
Use the skill only with consent from monitored people, confirm who receives alerts, and avoid submitting media or locations that should not leave the local environment.
If the dev configuration is selected, sensitive video or report data could be sent to an internal, unencrypted endpoint rather than the expected production service.
An included development configuration points to a private raw-IP HTTP endpoint; the provided artifacts do not prove it is active by default, but it could misroute sensitive uploads if dev mode is enabled.
base-url-open-api: "http://192.168.1.234:9601/smyx-open-api"
Verify the active config.yaml environment and use only trusted HTTPS production endpoints before running analysis.
Realtime monitoring may keep using camera/network resources and may continue producing alerts until the user stops it.
The realtime-stream analysis mode is designed to run continuously until manually stopped, which is expected for monitoring but important for users to understand.
print("按Ctrl+C停止监测")
...
while True:
ret, frame = cap.read()Run realtime monitoring only when intended, supervise it, and stop it explicitly when monitoring is no longer needed.