Back to skill

Security audit

Infant Blanket Kick Detection | 老年人夜间离床时长与徘徊识别

Security checks for vulnerabilities and agentic risk

Overview

This skill is for elderly night-monitoring video analysis, but it sends highly private bedroom footage and report queries through a cloud service while silently creating or reusing local identities and tokens.

Install only if you are comfortable with elderly bedroom or hallway footage, report metadata, and historical report queries being processed by the configured cloud service, and only with informed consent from the monitored person or their legal caregiver. Review who controls the backend account, where tokens are stored, whether reports are retained, and whether your deployment needs stricter confirmation or on-premise processing.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
Findings (62)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill claims focused resident-safety monitoring, yet also appears to perform external HTTP communication, authentication/token handling, identity generation, and workspace/directory creation. In combination, these undisclosed behaviors substantially expand the opportunity for sensitive video, report, and user metadata to be stored or transmitted beyond expected boundaries.

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-night-bed-exit-wandering-analysis"
description: "Using fixed cameras (infrared night vision) in nursing-home or home bedrooms, the system continuously monitors elderly bed-exit status and activity trajectory at night. | 通过养老院或居家卧室的固定摄像头(红外夜视),夜间连续监测老年人的离床状态和活动轨迹。输出异常预警,可联动护理人员手机或护士站大屏,防止老人走失、跌倒或发生意外。"
version: "1.0.17"
license: "MIT-0"
---

# 🛏️ Elderly Night Bed-Exit & Wandering Detection | 老年人夜间离床时长与徘徊识别
> **智能分析中枢** · �

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill handles extremely sensitive bedroom surveillance involving elderly individuals, yet the description does not clearly warn that media and reports may be transmitted to and stored in cloud services. This omission undermines informed consent, creates serious privacy/compliance exposure, and is more dangerous because the subject matter involves vulnerable individuals in private spaces.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Storing authentication-like tokens together with personal profile fields is not justified by the stated bed-exit monitoring use case and materially increases harm if the local database is accessed. In the context of elderly bedroom monitoring, unnecessary retention of tokens and PII creates a disproportionate privacy and account-compromise risk.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill declares broad operational behavior involving shell execution, file I/O, network access, and environment use, but does not constrain those capabilities with an explicit tool scope. In an agent setting, this increases the chance that the skill can access more resources than users or platform operators expect, especially when handling highly sensitive bedroom video and resident history data.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The AI role explicitly states 'Do not provide medical diagnosis or specific care operation plans, only output behavior statistics and alarm information.' Elsewhere, the skill says outputs include '建议与报告链接' and directs historical report listing/report-link rendering, which contradicts the narrower output promise in the role definition.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The default trigger is broad enough to auto-activate on generic video-analysis requests, which risks sending unrelated or excessively sensitive footage into this healthcare-oriented workflow without deliberate user intent. Overbroad activation is especially problematic here because the content may be intimate bedroom surveillance video.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill scope expands from media analysis into cloud history querying and report-link retrieval, which is a materially different capability involving access to previously stored sensitive records. In a bedroom-monitoring context, that broadening raises privacy risk because users may invoke analysis without understanding that the skill also supports retrospective data access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

Accepting a remote video URL without any explicit warning about upload, processing location, or data transmission is risky because bedroom monitoring footage is highly sensitive. In this context, users may unknowingly send elderly residents' nighttime video to remote infrastructure, creating privacy, consent, and regulatory exposure even if the transport itself is implemented elsewhere.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script's user-facing description consistently says it analyzes elderly nighttime bed-exit and wandering, yet it accepts --pet-type with values cat, dog, and other and passes that into ConstantEnum.DEFAULT__PET_TYPE. This is an active contradiction between the documented intent and the code's retained animal-classification interface, suggesting the implementation was repurposed from a pet-analysis skill without aligning behavior/docs.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The script resolves an internal user identity implicitly and hides the related CLI parameter from help output, which reduces user awareness and auditability. In a healthcare-adjacent monitoring context, undisclosed identity binding can lead to misattributed access, silent data association, or privacy violations when analysis results are fetched or stored under an unexpected account.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

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.

Content

Scanner excerpt · skills/smyx_analysis/scripts/skill.py (reported line 28)May include surrounding context.

python
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

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

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.

Content

Scanner excerpt · skills/smyx_analysis/scripts/skill.py (reported line 33)May include surrounding context.

python
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

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill accepts arbitrary HTTP(S) video URLs even though the declared use case is fixed-camera bedroom monitoring. This broadens the data ingestion surface to untrusted remote sources, can enable scope creep into arbitrary surveillance content, and may cause the backend analysis service to fetch attacker-controlled URLs or process unexpected content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

User-provided remote video URLs are submitted to the analysis service without any visible disclosure that external content references will be transmitted or fetched. In this surveillance context, that can expose sensitive source locations or cause users to unknowingly involve third-party resources in processing.

Content

No source excerpt is available for this finding.