Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly a cloud fall-detection client, but it silently handles identity, remote registration/login, token persistence, and sensitive media/report history in ways users should review before installing.
Install only if you are comfortable sending fall-detection images, videos, URLs, internal identity values, and report history to lifeemergence.com services, and with the skill creating a local workspace database that can store user identity and tokens. Avoid using it for highly sensitive care footage unless the publisher documents retention, deletion, authorization, and token storage practices clearly.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
Authenticated outbound requests, token/account management, workspace discovery, file access, and reading a local API-key-like file are all highly sensitive behaviors unrelated to the promised narrow function. In a skill handling home/nursing-home imagery, this broad access model creates substantial risk of credential misuse, privacy leakage, and unauthorized data movement.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "fall-detection-image-analysis"
description: "Detects whether anyone has fallen within a specified target area. Supports both image and short video analysis. Suitable for scenarios such as home care for elderly people living alone and safety monitoring in nursing homes. | 检测目标区域内是否有人跌倒，支持图片和短视频检测，适用于独居老人居家看护、养老院安全监测等场景"
version: "1.0.15"
license: "MIT-0"
---
# 🚨 Fall Detection & Analysis Skill | 跌倒检测分析技能
> **智能分析中枢** · 图片/视频智能分析 · 结构化报告 · 历史报告云端查询
---
##
The User model stores sensitive personal and authentication-related data, including token, open_token, email, birthday, age, and realname, none of which are clearly necessary for fall detection. In a safety-monitoring skill, collecting and updating such data broadens privacy risk, raises the impact of compromise, and may expose reusable credentials or session material.
The utility initializes and persists a default open-id, reads identity material from a local API-key file, and creates user records in a local database even though the advertised skill only performs fall-detection image/video analysis. This hidden identity bootstrapping expands the trust boundary, can silently bind actions to a synthetic user, and introduces undisclosed account/state manipulation unrelated to the stated function.
The HTTP helper performs login/registration via /sys/phoneLogin, populates authentication headers, retries authorization, and persists tokens locally—behavior far beyond image/video fall detection. In this skill context, covert account creation and token handling are especially risky because users would not reasonably expect a detection skill to authenticate to remote services or mutate account state on their behalf.
The skill advertises and instructs use of shell, network, file read/write, and environment-dependent behavior, but declares no explicit tool scope or permissions boundary. This increases the chance that an agent executes broader capabilities than users expect, especially given the skill also handles local files, URLs, cloud queries, and internal identity flows.
The auto-trigger phrases for historical report queries are broad enough to match ordinary conversation, causing unintended cloud queries and disclosure of report metadata. Because the reports concern elder-care monitoring, accidental triggering can expose sensitive history without sufficiently clear user intent.
The skill does not clearly warn users that uploaded images, videos, and supplied URLs may be transmitted to a cloud service for analysis. This is particularly dangerous given the likely sensitivity of in-home or nursing-home surveillance media, and users may not realize remote processing and retention implications.
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
Multiple user-facing strings are fixed in Chinese, such as the report header and export-link text, with no apparent opt-in or locale selection mechanism. This can violate language/locale policy when a skill forces a specific language on all users.
The manifest describes a fall-detection analysis skill for images and short videos, but the implementation also accepts any HTTP/HTTPS URL as input and forwards it for processing. Remote URL ingestion materially expands behavior beyond simple local media analysis by causing analysis of externally hosted content.
The skill reads the entire local file and uploads its contents to an external analysis service without any user-facing notice, consent step, or data handling disclosure in this code path. In a fall-detection context, those files may contain sensitive health, home, or elder-care imagery, so silent transmission creates a meaningful privacy and compliance risk.
The manifest claims a narrowly scoped capability: detecting whether someone has fallen, including support for both images and short videos. In this file, the main execution path only accepts video input or URL, delegates to a generic get_output_analysis, and also exposes show_analyze_list, with no indication of target-area handling, image support, or fall-detection-specific logic.
The analyze_video path passes either a local file path or URL to skill.get_output_analysis, indicating content may be transmitted off-host, yet the tool provides no prominent warning or consent flow. Because this skill targets fall detection in homes or nursing facilities, the media is likely highly sensitive, making undisclosed outbound transfer a significant privacy and compliance risk.
The script accepts a suppressed API key parameter and forwards analysis through an external skill/API path without clearly disclosing credential use or remote processing to the user. Hidden credential handling reduces transparency and can cause users to supply secrets or sensitive media under false assumptions about local-only operation.
Detected: suspicious.install_untrusted_source