Security audit
Security checks for vulnerabilities and agentic risk
The skill broadly matches a face-recognition cloud-analysis purpose, but it uses sensitive biometric media plus silent identity creation, login, and token persistence that are not scoped or explained clearly enough.
Review before installing. Use this only if you are comfortable sending face images/videos or camera URLs to the configured Lifeemergence backend, associating reports with a local or generated identity, and allowing local token/database persistence. Obtain consent from people shown in the media and avoid private or internal camera URLs unless you understand the service's retention and access policies.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The code characteristics indicate automatic account creation/login, token management, local file/database persistence, and generic backend HTTP access, none of which are clearly disclosed in the user-facing description. In a face-recognition skill, hidden authentication flows and persistent identifiers significantly elevate privacy, impersonation, and data-governance risk.
The YARA hit indicates suspicious metadata patterns in the manifest, which can be associated with tool or metadata poisoning attempts. Given the broader evidence of description-behavior mismatch and sensitive biometric scope, anomalous manifest content should be treated as a real trust-boundary issue because orchestrators may rely on metadata when deciding what the skill is allowed to do.
---
name: "familiar-person-recognition-analysis"
description: "Identifies acquaintances in videos or images through face photo comparison. Supports database enrollment, and the recognition results tell you who is at which location. Suitable for identity verification in homes and office areas. | 熟人识别分析技能，通过人脸图片比对识别视频/图片中的熟人，支持底库录入，识别结果告诉你哪个位置是谁，适用于家庭、办公区域身份核验"
version: "1.0.22"
license: "MIT-0"
---
# 👥 Acquaintance Recognition & Analysis Skill | 熟人识别分析技能
> **智能分析中枢** · 图片/视频智�
The description emphasizes convenience and precision but does not prominently warn that the skill handles biometric face data and may query cloud-hosted history reports. For a face-recognition workflow, insufficient privacy disclosure is dangerous because users may unknowingly submit highly sensitive personal data or retrieve identity-linked records without understanding storage and sharing implications.
The manifest says this skill identifies acquaintances in images or videos via face-photo comparison and supports enrollment, but this file exposes a generic 'video analysis' workflow and a history-listing command. There is no code here for face comparison, familiar-person recognition, image handling, location/person identity extraction, or enrollment/database management as described in the manifest.
This section reads identity values from local files, queries a local database for user records, and creates default open-id identities when none exist. For a familiar-person recognition skill, silent identity recovery and creation are out of scope and dangerous because they create durable user identities and backend linkage without transparency or necessity.
The utility layer silently performs identity resolution, login/registration, token acquisition, and persistence of user state, which is unrelated to a face-recognition analysis skill's declared purpose. In this context, hidden account provisioning and credential handling expand the skill's authority and enable unexpected tracking, impersonation, or backend account creation without informed user consent.
The code automatically performs network login/registration and stores returned tokens, all without an explicit warning or consent boundary. Combined with a face-recognition skill, this creates a serious transparency and privacy problem because biometric-related activity may become tied to remote accounts and persistent credentials behind the scenes.
The skill declares broad operational behavior including shell execution, local file handling, network access, and environment-dependent identity initialization, but provides no explicit permission or allowed-tools boundary. In an agent ecosystem, this increases the chance of over-privileged execution, hidden data access, and unintended command/network actions beyond what users would expect from a face-recognition skill.
The skill says it should trigger by default whenever a user needs to identify acquaintances in monitoring footage, which is broad and subjective. Vague auto-trigger rules can cause unintended biometric analysis or cloud report access without sufficiently explicit user intent, especially in multi-tool agent environments.
The instruction says internal identity parameters must not be shown, asked for, or required from the user, but the broader document is operationally centered on Chinese prompts and trigger phrases without explicitly offering a language or locale choice. This can amount to a locale/language policy issue because behavior appears optimized for a specific language without documented user opt-in.
This skill performs familiar-person recognition, which is biometric processing of highly sensitive personal data, on user-supplied images or videos without any notice, consent flow, retention disclosure, or usage limitation. In context, the skill is explicitly designed for identifying people in homes and offices, which increases privacy and surveillance risk because it can reveal who appears at specific locations and times.
The manifest describes identifying acquaintances in images/videos, supporting enrollment, and returning who is at which location. In this file, the code only validates an input file or URL, submits it to a generic analysis API, polls for completion, and formats/list reports; there is no recognizable logic for face comparison, enrollment, identity matching, or location extraction.
Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
result_json = JsonUtil.parse(result_json_pure_text, result_json_pure_text)
result_json_common_ai_response = result_json.get("commonAiResponse") if isinstance(result_json,
dict) else result_json
if result_json_common_ai_response:
result_json = result_json_common_ai_response
Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
result_json = JsonUtil.parse(result_json_pure_text, result_json_pure_text)
result_json_common_ai_response = result_json.get("commonAiResponse") if isinstance(result_json,
dict) else result_json
if result_json_common_ai_response:
result_json = result_json_common_ai_response
The skill forwards user-supplied remote video URLs to the backend service without any visible warning or confirmation. In this skill context, those URLs may reveal private camera feeds, internal resource locations, or personal surveillance content, so undisclosed transmission increases privacy exposure and can leak sensitive infrastructure details to the service provider.
Detected: suspicious.install_untrusted_source