Security audit
Security checks for vulnerabilities and agentic risk
The skill is a public-camera emotion-analysis tool that sends sensitive media to an external service and silently creates or reuses persistent account credentials.
Review this before installing. Only use it if you are comfortable sending public-camera images or videos to the configured external service, creating or reusing a persistent service identity, and storing associated tokens in the workspace database. Confirm consent, signage, retention, and privacy requirements for any public-place footage.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
The skill appears to perform authenticated external requests, local file operations, identity creation/retrieval, and persistent token/user storage despite being presented as anonymous crowd emotion analysis. In this context, undeclared identity and persistence features are especially dangerous because public-place video is highly sensitive and may involve biometric inferences.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "smyx-public-place-group-emotion-index-analysis"
description: "Using fixed cameras in malls, exhibition halls, scenic areas and other public places, the system analyzes facial expressions of multiple people in the scene in real time (with anonymized expression recognition only), aggregates the distribution of emotions (happy, calm, irritated, surprised, sad, fearful, etc.), and computes an overall group-emotion index (0-100; higher = more positive). | 通过商场、展览馆、景区等公共场所的固定摄像头，实时分析场景中多人的面部表情（使用匿名化表情识别），统计各类情绪（愉悦、平静、烦躁�
The description does not clearly warn users that local files or remote video URLs may be uploaded to an external/cloud API for analysis. Because the content involves public-place facial-expression footage, lack of explicit transfer disclosure significantly increases privacy, consent, and compliance risk.
The get_user_by_username method introduces user-identity lookup capability into a skill whose description emphasizes anonymous facial-expression analysis. That mismatch is especially concerning because it creates a path to correlate supposedly anonymized analytics with identifiable user records, undermining privacy expectations and expanding the system into identity-related data handling.
The skill manifest says the system performs anonymized group-emotion analysis in public places, but this module defines persistent user identity storage and mutation logic for sys_user records. That is a scope mismatch that expands the data footprint from anonymous analytics into identifiable account handling, increasing privacy, compliance, and insider-abuse risk without clear justification.
The User model stores username, realname, email, birthday, age, token, and open_token despite the manifest claiming anonymized facial-expression aggregation only. Token and identity storage materially increase the blast radius of compromise and are unjustified for a public-camera group-emotion skill, making the context substantially more dangerous from a privacy and surveillance perspective.
The code generates, persists, and reuses synthetic default user identities from local files and a database, despite the skill description emphasizing anonymized analysis. Persistent pseudo-identities can enable correlation across runs, silent account reuse, and user/activity tracking that exceeds the stated purpose.
The docstring frames the behavior as open-id initialization/default assignment, while the surrounding logic derives identities from a workspace file, database records, or generates new persistent usernames. In the context of a skill marketed as anonymous public emotion aggregation, this mismatch is dangerous because it conceals non-anonymous identity handling and undermines privacy expectations.
This utility performs token-backed login and implicit account provisioning that are unrelated to the manifest's claimed anonymous group-emotion analytics scope. That creates hidden identity establishment and outbound data flows, increasing the risk of undisclosed tracking, backend account creation, and transmission of user-linked identifiers to external services.
The skill advertises and operationalizes shell, file, network, and environment access without declaring any explicit tool scope or permissions boundary. This increases the risk of over-privileged execution, especially because the workflow instructs the agent to run local scripts, read/write files, and contact external services.
The default trigger activates on broadly defined public-place video analysis requests, which increases the chance that the skill runs unexpectedly on sensitive media. In a skill that uploads or processes surveillance-like footage, overbroad activation materially raises privacy and data-transfer risk.
The keyword trigger list includes broad and ambiguous terms that could match ordinary requests unrelated to this specialized surveillance-oriented analysis. That can cause unintended activation, unnecessary cloud uploads, or access to historical reports without sufficiently specific user intent.
The skill states that it will automatically create or reuse internal user identities, even though the advertised purpose is anonymous group emotion analysis. Hidden identity bootstrap and linkage create an unnecessary privacy risk and make historical activity easier to associate across sessions without user awareness.
This code hard-codes user-facing help text and runtime messages in Chinese, including the argparse description and error/progress output. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.
Detected: suspicious.install_untrusted_source