Security audit
Security checks for vulnerabilities and agentic risk
This skill performs the advertised reptile video analysis, but it silently creates or reuses a local cloud identity and stores account tokens/history context.
Review before installing. Use it only if you are comfortable sending reptile enclosure media or URLs to the configured LifeEmergence cloud service and allowing the skill to create/reuse a local identity database with stored tokens. Avoid use in shared workspaces unless account association and history access are clearly controlled.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
Reading local credential-like data, creating or resolving user identities, persisting them locally, and making authenticated outbound requests is a substantial capability escalation beyond reptile media analysis. In this context, the skill can handle credentials and user-linked cloud data without clear disclosure, creating risks of unauthorized access, data exfiltration, and account abuse.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "smyx-reptile-thermoregulation-behavior-analysis"
description: "Through fixed enclosure cameras, the system analyzes behavior videos of reptiles (lizards, snakes, turtles) and detects movement frequency and dwell duration between the basking zone (heated area under the basking lamp) and the hiding zone (cave/cool side). | 通过爬宠箱固定摄像头，分析爬行动物（如蜥蜴、蛇、龟）的行为视频，检测宠物在晒点（加热灯下方高温区域）与躲避区（洞穴、冷区）之间的移动频次、停留时长以及活动节律。系统连续监测，生成每日温区利用报告，异常时推送提醒。"
versio
The skill omits a clear warning that uploaded media and report queries may be transmitted to cloud services. For a video-analysis skill handling user media and identity-linked history, lack of disclosure undermines informed consent and can expose sensitive files or metadata unexpectedly.
Automatic identity initialization, fallback to a default local user, and silent user creation exceed the stated purpose and create hidden account/state management. In a multi-user or shared environment, this can misattribute data, expose another user's history, or create records without informed consent.
Using a local default user or creating a new user automatically is context-inappropriate for a reptile video-analysis skill and can cause cross-user data access or unauthorized report association. The danger is amplified because the skill also references cloud history retrieval, so a mistaken identity can fetch or store the wrong person's data.
This file includes generic user-account persistence logic that is unrelated to the declared purpose of reptile enclosure video behavior analysis. Capability overreach increases attack surface by introducing identity storage, account lookup, update, and deletion functionality that could process personal data without necessity or user expectation.
The User model stores identity attributes and especially token/open_token values, which are unjustified for a reptile thermoregulation video-analysis skill. If compromised or misused, these fields could expose authentication material and personal information unrelated to the advertised camera analytics function.
The code silently resolves, creates, and persists synthetic open-id/user identity state even when the user did not explicitly provide one. For a reptile enclosure video-analysis skill, this behavior is unrelated to core functionality and dangerous because it creates durable identity state that can later be used for remote requests and tracking without clear authorization.
This utility module performs broad authenticated network operations, including token handling, account lookup/provisioning, and arbitrary HTTP requests, which are unrelated to the stated reptile behavior-analysis function. In the context of a camera-analysis skill expected to work locally, hidden remote account and API activity materially expands the attack surface and enables undisclosed data exfiltration or backend abuse.
The skill declares broad operational behavior that implies shell, filesystem, environment, and network access, but it does not explicitly scope or restrict allowed tools. In an agent setting, missing tool-scope declarations increase the chance of over-privileged execution and make it harder to enforce least privilege for a skill that can access local files, identity state, and remote APIs.
The broad keyword trigger list can cause unintended activation for general reptile discussions rather than clear requests for analysis. Because this skill can save local files, call scripts, and query cloud history, accidental triggering can lead to unnecessary data processing or unexpected external requests.
The documentation expands the skill from analysis into cloud historical report retrieval and report-link output, increasing external data access beyond the headline purpose. This is risky because users may not expect prior records and report URLs to be queried or exposed when invoking a behavior-analysis skill.
The documentation contradicts itself by claiming cloud-only history retrieval while also describing local default-user reuse and local user creation. These inconsistencies are dangerous because they obscure actual trust boundaries and make it difficult to reason about whether report association is secure and correctly scoped.
The script exposes a history-listing path tied to an open_id, which extends beyond simple one-shot video analysis into account-scoped data access. In this file, there is no visible authorization check, user disclosure, or scope restriction before retrieving historical reports, creating a risk of unintended access to prior analysis data if identity handling elsewhere is weak or spoofable.
The code initializes an internal account identity via OpenIdUtil and accepts a hidden --open-id argument suppressed from help, which is not justified by the stated reptile video-analysis purpose. Hidden identity plumbing increases the chance of stealthy account-context operations and can enable unauthorized data access or cross-user confusion if callers can influence the identity value.
Detected: suspicious.install_untrusted_source