Back to skill

Security audit

Elderly Gait Instability / Shuffling Step Detection | 老年人步态不稳/小碎步识别

Security checks for vulnerabilities and agentic risk

Overview

This health-video skill is not clearly malicious, but it silently links analyses to persistent identities and cloud report history, so it needs review before installation.

Install only if you are comfortable sending elderly gait videos or URLs to the configured cloud service and having reports tied to a persistent internal identity. Before using it with real health footage, require informed consent from the person or guardian, confirm the publisher's retention/access-control policy, and prefer explicit user confirmation before report-history lookup or export-link generation.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (62)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill appears to perform external HTTP requests, manage auth headers/tokens, resolve user identities, and read/write local workspace data, none of which are clearly reflected in the top-level purpose. This is especially dangerous because the inputs are health-related videos and historical reports, making undisclosed data transfer and identity linkage higher-risk than in a generic media tool.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: "smyx-elderly-gait-instability-detection-analysis"
description: "Using a fixed camera in a hallway or living room to record video of an elderly person walking in a straight line, AI pose estimation and gait analysis extract parameters such as step length (cm), gait speed (m/s), trunk sway angle (left-right tilt), and cadence to evaluate gait stability. | 通过走廊或客厅的固定摄像头拍摄老年人直线行走的视频,利用AI姿态估计和步态分析技术检测步幅长度(cm)、步速(m/s)、躯干摇摆角度(左右倾斜度)以及步频等参数,评估步态稳定性。当步幅过小(小碎步)、步速�

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

Automatic creation or reuse of internal user identities is not obviously necessary for local gait analysis and materially increases the privacy and account-linkage risk. In this context, silent identity binding can connect sensitive health-video analysis with persistent profiles without meaningful user awareness.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
91% confidence
Finding

The model stores identity data and authentication-style secrets (token, open_token) in a local SQLite database without visible encryption, hashing, retention controls, or access restrictions. In the context of an elderly gait-analysis skill, collecting and persisting such data appears weakly justified and increases the risk of credential leakage, privacy harm, and cross-context account correlation if the local database is accessed.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This utility file contains identity resolution, credential loading, token management, auto-registration/login, and authenticated remote request logic that is unrelated to the declared local gait-analysis purpose. In the context of a health-oriented video analysis skill, this hidden account and network behavior materially expands the attack surface and can silently move user identity and telemetry off-device.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code reads an internal identity value from a workspace file and, if absent, creates or reuses a default user identity in a local database. For a gait instability detection skill, silent identity discovery and account bootstrapping are unrelated to the stated functionality and can enable tracking, impersonation, or unauthorized linkage of health-related activity to persistent identifiers.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
85% confidence
Finding

The skill declares no explicit tool/permission scope while its documented behavior includes shell execution, local file handling, environment access, and network/API interaction. In an agent environment, missing scope boundaries increases the chance of unintended command execution or data access beyond the user’s expectation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill claims to only output objective metrics and risk grading, but elsewhere expands to medical review/rehabilitation advice and cloud report-link delivery. This inconsistency can mislead users and orchestrating agents about the skill’s authority and data-handling scope.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The role section says not to provide medical diagnosis or clinical advice, yet the capability list includes medical review/rehabilitation recommendations. In a healthcare-adjacent context, contradictory instructions can cause unsafe reliance, inappropriate advice, or policy bypass by downstream agents.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The default trigger is broad enough to analyze any uploaded elderly walking video without clear affirmative intent. For sensitive health-related media, accidental invocation can lead to unnecessary processing, storage, or transmission of personal data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The keyword triggers include broad phrases such as unstable walking or fall risk that can appear in ordinary conversation. This raises the chance of unintended activation and processing of sensitive video or health-related context.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill broadens from one-time gait-video analysis to automatic retrieval of historical reports tied to internal identity. That expansion increases privacy risk because users may intend to analyze a file but not authorize account-linked health-history access.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Automatic history-report query based on broad natural-language phrases can cause identity-linked cloud retrieval when the user may only be speaking generally. In a health-monitoring setting, unintended access to prior reports is a meaningful privacy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The method returns a fixed Chinese string for the analysis output header with no indication that the user can choose language or locale. This can violate a policy requiring language choice or opt-in, since the skill appears to force a specific language in user-facing output.

Content

No source excerpt is available for this finding.