Security audit
Security checks for vulnerabilities and agentic risk
This skill sends pet media to a cloud service and automatically creates or reuses a local identity with persisted tokens, which is broader than its pet-detection label makes clear.
Install only if you are comfortable with pet images/videos or URLs being sent to the lifeemergence cloud service, automatic account creation or reuse, local token storage in the workspace data directory, and cloud history/report retrieval. Avoid using it with sensitive home footage unless you can verify retention, account deletion, and token cleanup controls.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
Remote API access, automatic registration/login, filesystem workspace creation, local file writes for identity material, and local user persistence create a much broader and more sensitive behavior profile than pet detection alone. This is dangerous because it combines network exfiltration, identity persistence, and local state mutation, substantially increasing privacy and compromise risks.
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
---
name: "pet-detection-analysis"
description: "Detects cats, dogs, and birds appearing in the target area; supports video stream and image detection, suitable for home pet monitoring scenarios. | 宠物检测技能，检测出目标区域内出现的猫、狗、鸟，支持视频流和图片检测，适用于家庭宠物监控场景"
version: "1.0.20"
license: "MIT-0"
---
# 🐾 Pet Detection Skill | 宠物检测技能
> **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
---
## 🧭 技能概览 | Overview
| 模块 | 内容 |
|---|---|
| 🏷️ 技能名称 | **宠物检测
The documented APIs are for pet health analysis and report export, while the skill metadata claims pet detection for cats, dogs, and birds. This mismatch can cause an agent or operator to invoke unrelated health-analysis functions, potentially exposing sensitive health-report workflows and data paths that were not intended for this skill.
The referenced endpoints and scenario code indicate report generation, result retrieval, and export behavior for pet health analysis rather than object detection. In an agent setting, this kind of capability confusion is dangerous because it may grant access to sensitive historical reports or exports under the guise of a benign pet-monitoring skill.
The implementation is a generic analysis/reporting client rather than a narrowly scoped pet-detection skill. It handles arbitrary analysis responses, report export links, and polling/report workflows with no code enforcing cat/dog/bird-only behavior, creating a scope mismatch that can expose broader backend capabilities than users expect.
This file exposes a generic API service layer with broad remote capabilities (add, edit, delete, arbitrary GET/POST/PUT/DELETE, pagination helpers, and user lookup) that materially exceeds the declared scope of a pet-detection skill. In a skill meant for local image/video pet recognition, such unrestricted network primitives increase the attack surface and can enable unauthorized data access, exfiltration, or remote state changes if other parts of the skill invoke them with attacker-controlled inputs.
The http_post/http_put/http_get/http_delete methods allow arbitrary outbound requests to caller-supplied URLs, which is not justified by the stated pet-monitoring purpose. In this context, these wrappers can be abused for SSRF-style access to internal services, data exfiltration, or using the skill runtime as a generic proxy, making the mismatch between capability and declared function especially dangerous.
This file defines a generic ai_chat interface inside a skill advertised for pet detection, creating a capability boundary mismatch. Even though the subprocess execution is currently commented out and the function is broken, the presence of an unrelated agent-invocation pathway in a narrowly scoped monitoring skill is dangerous because such code can later be enabled to send arbitrary prompts to an external agent, expanding the skill into a general-purpose execution or exfiltration surface.
The code introduces a general agent-execution capability that is unjustified by the stated home pet monitoring purpose. In context, that mismatch is especially risky because users or reviewers may grant this skill trust and access appropriate for image/video detection, while hidden or surplus agent functionality could later be repurposed for arbitrary prompt handling, data leakage, or command execution if the commented subprocess path is restored.
The utility layer performs remote account provisioning, token acquisition, local token persistence, and authenticated API bootstrapping that are unrelated to the stated pet-detection purpose. In a skill expected to process images/video, hidden identity creation and credential management materially expand scope and can cause silent outbound data exchange and account-side effects without informed user consent.
The skill documentation instructs use of shell execution, network access, local file handling, and implicit identity/config storage, but it declares no explicit tool scope or permission boundaries. This is dangerous because an agent may overgrant capabilities to the skill, making unintended file, shell, or network actions easier to trigger without clear review.
Default activation on any uploaded media or URL plus vague mention of analysis intent is too broad for a skill with networked processing and identity linkage. This is dangerous because users may trigger remote upload, storage, or report creation without clearly opting into this specific skill or understanding the consequences.
Cloud-based historical report querying and persistent user association are secondary data-processing capabilities not necessary for basic pet detection. This is dangerous because it expands the data lifecycle and allows cross-session aggregation of user activity and media-derived reports, increasing privacy and unauthorized-access risk.
The history-query trigger phrases are broad enough to activate cloud report retrieval on common language that may not indicate informed user intent. This is dangerous because it can cause unintended access to historical data or backend calls, especially in a multi-user or privacy-sensitive environment.
Automatic creation and reuse of a local default identity is unrelated to core media inference and silently establishes persistent user state. This is dangerous because it can associate future analyses to the wrong user, create hidden tracking identifiers, and leave residual identity artifacts on the local system.
Detected: suspicious.install_untrusted_source