Security audit
Security checks for vulnerabilities and agentic risk
The skill does gait-analysis work, but it also silently links users to cloud accounts, stores tokens locally, and can fetch sensitive history reports with limited user control.
Install only if you are comfortable sending elderly gait videos or private video URLs to the Life Emergence cloud service and allowing the skill to create/reuse a local identity with stored tokens for report history. Use explicit consent from the recorded person or caregiver, avoid sensitive signed/private URLs unless intended, and review account/history access controls before production use.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The documented implementation reads/writes local paths, generates and persists user/open-id identities, accesses a local user DB, performs external authentication, manages tokens, and issues generic API requests. For a health-related media-analysis skill, this undisclosed identity and backend access chain materially increases privacy, authorization, and data-exfiltration risk.
The YARA match indicates metadata-poisoning indicators in the manifest description, such as suspicious hidden or malformed characters. While this may be accidental, such obfuscation can be used to evade scanners, alter parsing behavior, or hide misleading metadata, which is especially concerning when combined with other description/behavior mismatches.
---
name: "smyx-elderly-gait-instability-detection-analysis"
description: "Using a fixed camera in a hallway or living room to record video of an elderly person walking in a straight line, AI pose estimation and gait analysis extract parameters such as step length (cm), gait speed (m/s), trunk sway angle (left-right tilt), and cadence to evaluate gait stability. | 通过走廊或客厅的固定摄像头拍摄老年人直线行走的视频，利用AI姿态估计和步态分析技术检测步幅长度（cm）、步速（m/s）、躯干摇摆角度（左右倾斜度）以及步频等参数，评估步态稳定性。当步幅过小（小碎步）、步速�
Automatic creation and reuse of a local default user is an unjustified identity-management capability for this skill and can cause cross-session or cross-user data mixing. If exploited or misconfigured, historical reports and health-related outputs may be associated with the wrong identity or exposed to unintended parties.
The skill is described as elderly gait-instability analysis, but this module defines a generic sys_user persistence layer including identity fields and token storage. That data collection and retention materially exceeds the stated purpose, increasing privacy, compliance, and breach exposure without clear necessity for the gait-analysis workflow.
The User model stores authentication-style tokens along with personal profile data such as real name, email, birthday, sex, and age, yet the declared skill function is video-based gait-risk scoring. In this context, collecting and persisting sensitive identifiers and tokens is disproportionate and dangerous because compromise of the local database could expose both PII and reusable credentials.
This utility contains account provisioning and remote login logic that is unrelated to elderly gait analysis, including silent registration/login behavior and token acquisition. In the context of a healthcare-oriented video analysis skill, hidden identity bootstrapping and service authentication expand the data-handling surface and can enable undisclosed outbound account creation, user impersonation, or backend access without clear user consent.
The skill declares no explicit tool scope or permissions even though the documentation indicates capabilities spanning shell execution, network access, filesystem access, environment use, and local data persistence. This under-specification weakens containment and review, making it easier for the skill to exercise broader capabilities than users or the platform may expect.
The skill description does not clearly warn users that URL inputs and history queries send data to cloud/API services. With elderly gait videos and fall-risk reports constituting sensitive personal and health-adjacent information, lack of upfront disclosure undermines informed consent and increases privacy risk.
The AI role explicitly restricts output to objective gait indicators and risk grading, excluding medical diagnosis or clinical advice. Later, the capability table states the skill can provide '医疗复核/康复建议' (medical review/rehabilitation recommendations), which directly conflicts with that restriction.
The default-trigger rules are broad and can cause the skill to activate on common elderly-walking analysis requests without tight scoping. Over-broad triggering is risky because it may automatically invoke networked analysis and sensitive health-data handling more often than the user intended.
The documentation expands the feature set from single-video gait analysis into cloud-hosted historical-report retrieval, alerting, and report-link workflows without reflecting that broader scope in the core description. This creates a transparency gap around storage, secondary use of data, and access to prior health-related reports.
Broad natural-language triggers for history-report queries can auto-invoke cloud retrieval of prior reports with insufficient scope constraints. In this context, that means potentially exposing or fetching sensitive historical health-related data based on casual phrasing alone.
Mandatory cloud API querying for historical reports grants broader data-access capability than required for one-off gait analysis. Because the data concerns elderly mobility and possible health conditions, unnecessary report-list access meaningfully increases privacy sensitivity and attack surface.
Automatic identity initialization and reuse of a local default user go beyond the stated task of gait analysis and introduce hidden user-state management. In a sensitive health context, undisclosed identity reuse can link sessions and reports in ways the user does not expect.
Detected: suspicious.install_untrusted_source