Security audit
Security checks for vulnerabilities and agentic risk
This skill is not clearly malicious, but it quietly combines cloud media submission, account creation, and token persistence with an overstated snake-health image-analysis description.
Review before installing. Use it only if you are comfortable sending snake images/videos or URLs to the configured LifeEmergence cloud services and having a local workspace database store account/session tokens. Clarify retention, deletion, and account linkage with the publisher, and do not treat its output as a veterinary diagnosis.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The strongest mismatch finding indicates authenticated external API communication, local file/directory management, identity resolution, token handling, and even payment/balance response handling, while no actual image-analysis logic is present. That combination is dangerous because it disguises a broad data-processing and account-linked client under a narrow animal-health skill description.
The YARA hit for metadata poisoning indicators is plausible because the manifest contains unusually dense, persuasive behavioral metadata and the overall file strongly overstates capability relative to implementation. While the exact signature basis is opaque, in context this can be treated as suspicious manifest-level prompt/tool steering that may mislead agents about what actions to take.
---
name: "smyx-snake-stomatitis-detection-analysis"
description: "Through fixed enclosure cameras, the system captures high-definition images of the moment a snake opens its mouth (yawning, post-feeding, or oral examination) and uses AI visual analysis to detect oral mucosa color (normal pink, mild inflammation red, severe inflammation dark-red or pale), the presence of pus points (white or yellow dots), ulcers, or necrotic tissue (irregular depressions, necrotic patches), comprehensively. | 通过蛇箱固定摄像头，捕捉蛇张口（打哈欠、进食后或口腔检查）时的瞬间高清图像，利用 AI 视觉分析技术检测口腔�
The implementation materially diverges from the declared skill purpose: instead of targeted snake stomatitis image analysis, it exposes a generic video analysis/history workflow via skill.get_output_analysis and skill.get_output_analysis_list. This kind of capability mismatch is dangerous because it can hide broader data handling behavior than users expect, increasing the risk of unauthorized collection, transmission, or repurposing of media under a misleading medical/animal-care description.
This file exposes a broad generic API client surface including add/edit/delete and raw http_get/http_post/http_put/http_delete methods, which goes well beyond the stated snake oral-image analysis purpose. In an agent/skill environment, such unrestricted network and CRUD primitives can be repurposed by other code paths to access unrelated internal or external services, exfiltrate data, or perform unauthorized state-changing actions.
The User model stores personal profile data and sensitive authentication material such as token and open_token, which is unjustified for a snake oral-condition detection skill. Collecting and retaining such data without an evident functional need creates unnecessary compromise risk and significantly worsens the severity of any local data exposure.
The utility implements identity resolution, reads local workspace identity material, queries a local user database, and auto-creates default users unrelated to the advertised snake image-analysis function. This expands the skill's capabilities into account management and identity persistence without clear necessity or disclosure, increasing the risk of unauthorized account linkage, privacy violations, and misuse of local agent state.
The HTTP layer performs broad external authentication and token-handling behavior, including auto-login/registration and token persistence/retry, which is materially beyond the stated oral-disease vision-analysis behavior. Because this code can transmit identity data and manage session credentials automatically, it creates a hidden data-flow and trust boundary crossing that users would not reasonably expect from the skill description.
The skill declares no explicit tool restrictions even though its documented behavior requires shell execution, local file access, environment access, and network connectivity. In an agent environment, missing scope declarations can lead to overbroad execution privileges and make it easier for the skill to access or exfiltrate data beyond what users would reasonably expect.
The skill description does not clearly warn users that local files and remote URLs may be sent to external/cloud services for analysis and history lookup. This is a real transparency and privacy issue because users may assume local-only veterinary image analysis from the description.
The documentation claims automatic fixed-camera capture, but the workflow actually depends on user-supplied files or URLs. This inconsistency can mislead users about what data is collected automatically versus what they are manually submitting, weakening informed consent and auditability.
Claiming automatic mouth-open detection and triggered snapshots when only offline script invocation is documented is another material misrepresentation of system behavior. In aggregate with the other mismatches, this increases the risk of overbroad trust and misunderstanding of data collection boundaries.
An overly broad default trigger may cause the skill to activate for any snake mouth image/video request, even when the user did not intend cloud submission or this specific workflow. In an agent setting, over-triggering can lead to unexpected file handling, network transmission, and privacy-impacting actions.
Automatic history-query triggers based on broad natural-language phrases can cause identity-linked cloud lookups without sufficiently specific user authorization. This is especially sensitive because the skill also associates reports with internal user identity behind the scenes.
Mandating cloud API access and local default-user creation is not obviously necessary for the stated purpose of analyzing uploaded snake-mouth images. Hidden identity creation and required cloud linkage can enable tracking, unexpected persistence, and privacy loss beyond user expectations.