Security audit
Security checks for vulnerabilities and agentic risk
This skill appears to be a cloud-backed infant cry analysis tool, but it silently creates or reuses identities, stores tokens locally, and has unclear media and privacy boundaries for sensitive child recordings.
Review this skill before installing. Use it only if you accept sending infant audio/video and report history to the lifeemergence.com cloud service, and verify that guardians consent, retention/deletion terms are acceptable, and local token storage in the workspace data directory is appropriate.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
The finding indicates undeclared remote authentication/token management and local file/persistence interactions, none of which are disclosed in the user-facing description. In a skill processing infant audio/video, undisclosed auth and storage behaviors materially increase privacy, consent, and potential account-linkage risks.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "smyx-infant-cry-cause-classification-analysis"
description: "Using the built-in microphone of a baby monitor or smart camera to capture infant cry audio, AI acoustic analysis extracts cry features such as frequency, pitch, rhythm, and duration, and classifies the possible causes behind the cry (hunger, sleepiness, pain/discomfort, boredom/need for comfort, fear, etc.), outputting the most likely cause and its confidence. | 通过婴儿监护器或智能摄像头的内置麦克风采集婴儿哭声音频，利用AI声学分析技术提取哭声的频率、音调、节奏、持续时间等特征，分类识别婴儿哭声背后的可能原�
The skill does not clearly warn users that uploaded audio/video files and URLs may be transmitted to cloud APIs for processing. Because the data concerns infants and potentially includes in-home recordings, the lack of prominent disclosure creates a serious consent and privacy risk and makes the skill more dangerous than a typical media-processing tool.
The manifest describes a narrowly scoped capability: capturing infant cry audio and classifying likely cry causes from acoustic features. This code explicitly requires a local video path or network video URL, sets a videoUrl parameter, and later uploads file content for analysis, indicating broader video-analysis behavior than the stated audio-only cry classification purpose.
The implementation materially diverges from the declared skill purpose: it analyzes video inputs and exposes video-history behavior instead of performing infant cry audio classification. This is dangerous because users and platform reviewers may grant microphone/child-monitor trust based on the manifest, while the code actually processes different media and can route arbitrary local paths or remote URLs into undisclosed backend analysis functionality.
This utility module implements identity resolution, local user creation, token caching, authenticated API calls, and account lifecycle behavior that materially exceeds the declared infant-cry classification purpose. In the context of a baby-monitor skill, hidden generalized account provisioning and tokenized backend access create an unjustified capability expansion that can transmit identifiers, act on behalf of users, and access unrelated services without clear consent.
The code automatically posts user identifiers to an external '/sys/phoneLogin' health endpoint with registration enabled, silently creating or retrieving accounts. For an infant-cry classifier, this behavior is unrelated to the advertised function and exposes users to undisclosed identity transmission, backend account creation, and service linkage without informed consent.
The skill declares no explicit tool/permission boundaries even though the documented workflow requires shell execution, network access, file reads/writes, and environment use. In an agent setting, missing scope declarations can let the skill exercise broader capabilities than users or policy expect, increasing the blast radius if the skill is misused or the backing scripts are unsafe.
The manifest mixes infant cry audio analysis with image/video intelligent analysis and broad media input types, creating ambiguity about what content the skill will accept and process. Ambiguous modality declarations increase the chance of unintended data collection and misuse of unrelated media, especially where family or nursery footage may be involved.
The addition of cloud historical report query and report-link behavior goes beyond the primary real-time analysis description and introduces secondary data-access capabilities. Secondary capabilities are risky when not foregrounded, because users may not realize that prior reports are stored remotely and retrievable via linked identity.
The AI role says not to provide medical diagnosis, yet other parts of the skill reportedly use diagnosis-style language and report framing. In a baby-care context, this inconsistency can encourage users to overinterpret outputs as medical advice, creating safety and liability concerns if parents delay appropriate care.
A broad default trigger that activates whenever infant cry audio/video is supplied can cause the skill to run without sufficiently explicit user intent. In this context, that means sensitive recordings of minors could be processed or uploaded unexpectedly, which is a meaningful privacy and consent issue.
The infant-cry skill exposes a pet-type control that is unrelated to the stated functionality, and it mutates a shared default classification setting based on user input. This kind of hidden cross-domain behavior increases the risk of confused-deputy issues, unexpected backend routing, or abuse of code paths not intended for infant-audio analysis.
The skill includes a list operation that retrieves prior analyses by internal user identity, which goes beyond the manifest's described single-input cry classification behavior. This expands the data-access surface and may expose historical analysis records or metadata to callers who only expected one-off local inference behavior.
Detected: suspicious.install_untrusted_source