Security audit
Security checks for vulnerabilities and agentic risk
This pet-grooming analysis skill sends media to cloud APIs and silently creates or reuses an account identity with locally cached tokens, which is broader than users would reasonably expect.
Review this skill carefully before installing. It is not just an offline grooming-photo analyzer: it can send images/videos or URLs to lifeemergence cloud services, query cloud history, silently create or reuse an internal account identity, and cache returned tokens in a local workspace database. Install only if you accept those account-linking and remote-processing behaviors.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
The reported behavior includes authenticated API requests unrelated to pet analysis, local file/directory creation, workspace/runtime detection, identity generation, token management, local database interaction, and remote account provisioning. In the context of a harmless-seeming pet grooming tool, that breadth strongly suggests hidden data collection or platform abuse potential.
A metadata-poisoning indicator in the manifest is a serious red flag because tool/skill metadata is part of the trust boundary used by agents and reviewers to decide what a skill does. Combined with the many description-behavior mismatches here, suspicious metadata increases the likelihood that the package is designed to influence routing, trust, or permission decisions under misleading pretenses.
---
name: "smyx-grooming-effect-assessment-analysis"
description: "AI-powered pet grooming effect assessment: detects mat residue area, dandruff coverage, and coat smoothness from post-grooming images, outputs a 0-100 grooming score with targeted re-grooming suggestions. Scenarios: daily home grooming, pet salon quality check, long-hair cat/dog shedding season management. | 通过智能梳毛器或普通摄像头拍摄梳毛后的宠物皮肤和毛发高清图像，利用AI图像识别技术检测毛结团块残留面积、皮屑覆盖率以及毛发顺滑度，自动评估本次梳毛效果，并提示是否需要进行二次梳理或进一步�
The implementation accepts both local files and arbitrary HTTP(S) video URLs, despite the skill being described as post-grooming image assessment from still images. This capability mismatch expands data collection and processing beyond the documented purpose, increasing privacy, consent, and misuse risk because users may unknowingly submit broader or more sensitive media than intended.
The skill metadata describes post-grooming image assessment, but the implementation is a video-analysis CLI that also exposes a history-listing path via --list and hidden open-id handling. This semantic mismatch is dangerous because users, reviewers, and policy controls may approve or invoke the skill under false assumptions, while the code actually processes different inputs and can access prior analysis records, expanding data exposure beyond the declared purpose.
The file contains account provisioning, token handling, local identity persistence, and remote login/request orchestration that are unrelated to the declared pet grooming image-assessment purpose. In this context, the mismatch is dangerous because a user enabling an image-analysis skill would not reasonably expect hidden identity management and networked account actions, creating unnecessary credential exposure and covert capability expansion.
This code can silently generate a default open-id, create a corresponding platform user, and persist that identity locally. For a grooming assessment skill, automatic account creation is unjustified and can bind the user environment to an external service without informed consent, enabling tracking, billing linkage, or later token misuse.
The skill declares broad behavior that implies shell execution, local file access, environment access, and network/API communication, but it does not define any explicit tool scope or allowed-tools boundary. In an agent environment, this increases the chance of over-privileged execution and makes it harder to enforce least privilege or review what the skill is actually permitted to do.
The manifest presents the skill as a grooming-effect analyzer, but the documentation expands scope to cloud-backed historical report lookup and report-link retrieval. Scope expansion matters because it introduces remote data access and potentially account-linked records that users may not expect from a simple media-analysis tool.
Overly broad trigger phrases for history-report retrieval can cause the skill to invoke cloud report lookup on vague requests without sufficient user intent confirmation. In a skill tied to historical records, this increases the risk of accidental disclosure of prior reports or unnecessary remote data access.
The skill states that uploaded attachments are automatically saved locally without clearly warning users about local data retention. Because the content consists of user-supplied pet images/videos, undisclosed retention creates a privacy and data-governance issue, especially if the files contain metadata or identifiable household context.
The skill requires direct cloud API queries for historical reports but does not clearly tell users that report data will be fetched from a remote service. This lack of transparency can result in unexpected third-party data transfer and undermines informed consent for history retrieval.
The skill includes a history listing function that returns prior analysis reports by open_id, which goes beyond the manifest's stated purpose of analyzing grooming images. Hidden or undocumented access to historical reports can expose sensitive prior user activity or pet-related records, especially when combined with implicit identity resolution and suppressed CLI arguments.
The script silently resolves an internal user identity via OpenIdUtil.resolve_current_open_id even though pet grooming image analysis does not require hidden identity handling. This creates unnecessary user tracking and enables backend actions to be tied to an internal account context without explicit user awareness or consent, which is especially concerning because the same script also exposes history retrieval functionality.
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
Detected: suspicious.install_untrusted_source