Security audit
Security checks for vulnerabilities and agentic risk
The skill is not clearly malicious, but it handles highly sensitive child bedroom audio/video through cloud APIs while silently creating and persisting user identity tokens, so it needs careful review before installation.
Install only if you intend to send child nighttime audio/video or report history to the configured cloud service and accept silent account/token creation in a local workspace database. Confirm guardian consent, retention/deletion terms, and access controls before using real children's bedroom recordings.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
Authenticated HTTP communication, local file and API-key handling, generated user identities, and remote auto-registration/login are materially different from the declared child sleep analysis use case. These hidden behaviors can lead to silent account creation, credential exposure, and transmission of children's bedroom media and metadata to external systems without adequate transparency.
The metadata poisoning indicator suggests malformed or suspicious manifest metadata that could interfere with tool interpretation or conceal actual behavior. While this may be accidental formatting, in a skill manifest it can still be used to obscure permissions, descriptions, or downstream parsing in ways that reduce security review effectiveness.
---
name: "smyx-child-nightmare-rollover-detection-analysis"
description: "Using a fixed camera in the child's bedroom (infrared night vision), the system continuously captures video and audio at night to analyze the child's sleep behavior. It detects rollover frequency (rollovers per minute), cries (recognizing specific cry-sound features), and sleep talk (speech during sleep), and generates a sleep-quality report. When rollovers occur too often (e.g., > 3 per hour), strong crying is detected, or sleep talk is observed, the system pushes 'possible nightmare' or 'restless sleep' alerts to the parents. Application scenarios: child bedrooms, infant rooms.
The description does not clearly warn users that highly sensitive children's bedroom audio/video and related historical reports may be handled through cloud-facing services. In this context, missing disclosure is particularly dangerous because minors' intimate in-home recordings carry elevated privacy, consent, and compliance obligations.
The User model stores authentication tokens and broad personal profile data such as real name, email, birthday, sex, and age, none of which are clearly necessary for detecting nightmares or restless sleep. In a system monitoring children in bedrooms, unnecessary retention of identity and auth secrets materially raises harm from compromise, correlation, and misuse.
This utility layer automatically provisions or logs in users against a remote service, retrieves tokens, and persists them locally, even though the declared skill purpose is child sleep/nightmare monitoring. That creates undisclosed account creation and identity binding behavior, increasing privacy risk and enabling external data linkage unrelated to the stated function.
The skill advertises and instructs use of capabilities equivalent to shell execution, local file access, network access, and environment use, but declares no explicit tool-scope or permission boundary. In an agent setting this weakens least-privilege controls and can allow the skill to access sensitive local media, invoke scripts, and contact remote services without transparent restriction.
The skill claims current sleep-media analysis, but also includes cloud historical report querying and link retrieval beyond the stated monitoring scope. This scope expansion matters because it enables access to prior sensitive child-room recordings and reports, not just analysis of the file the user provided now.
The AI role says the skill should not provide medical or sleep-disorder diagnosis, yet trigger keywords include requests for 'sleep health diagnosis reports.' This inconsistency can cause unsafe reliance on the tool for health conclusions about a child and increase legal and safety risk.
The automatic trigger conditions are broad enough to activate on common sleep-related requests, potentially causing unintended processing of sensitive child audio/video or unnecessary cloud history queries. Over-broad invocation increases the chance of privacy-invasive actions without clear, informed user intent.
This documentation describes continuous overnight audio and video monitoring of a child, including cry and sleep-talk detection, which involves highly sensitive biometric and behavioral data. Referencing only API key authentication and omitting explicit privacy, consent, retention, access-control, and sensitive-data handling requirements creates a real security/privacy weakness because implementers may deploy or integrate the system without safeguards appropriate for children's data.
The string returned to users is fixed as Chinese text, which can impose a specific language/locale on all users. The file does not show any opt-in, language selection, or justification that this skill is intentionally limited to Chinese-speaking users.
The manifest is specifically about analyzing a child's nighttime sleep behavior, yet the CLI and analysis wrapper accept a pet_type parameter with values cat, dog, and other, and write it into ConstantEnum.DEFAULT__PET_TYPE. This indicates the implementation is partially repurposed from a pet-analysis skill and exposes behavior outside the manifest's stated child-sleep scope.
The manifest promises specific behaviors such as detecting excessive rollovers, strong crying, or sleep talk and pushing alerts to parents. In this file, the code only calls skill.get_output_analysis(...) and skill.get_output_analysis_list(...) and provides no visible logic for threshold evaluation, alert generation, or parent notification, creating a mismatch between the described behavior and the implemented operations in this entrypoint.
This Python code includes command descriptions and help text entirely in Chinese, which imposes a specific language on users. Under the policy, locale-specific behavior should either provide a user choice or clearly document a justified locale constraint; neither is present in this file.
Detected: suspicious.install_untrusted_source