Back to skill

Security audit

Real-time Employee Absence Monitoring Skill | 人员离岗实时监测技能

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious with high confidence because scanners report credential and account handling, video uploads, and behaviors that do not clearly match its advertised personnel-monitoring purpose.

The verdict is suspicious, not malicious. Before installing, consider whether the account registration, local token storage, video uploads, and report-history access fit your needs and privacy requirements; the scan is incomplete because one local reference was missing.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (26)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk implements generic user-data persistence and default-user lookup. It contains no camera processing, pose estimation, area monitoring, absence detection, threshold evaluation, or alert generation. Database persistence could support the declared skill, but this chunk specifically manages identities and tokens without an evident connection to monitoring configuration or results. The mismatch concerns the supplied code only; it does not establish what other components of the skill implement. No network transmission or token exfiltration is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

Ordinary networking, polling, and file helpers could support the declared monitoring function and are not independently evidence of a mismatch. However, this code also implements an undisclosed account lifecycle and billing integration: /sys/phoneLogin receives silent=1 and register=1, returned credentials are persisted, and HTTP 402 responses produce payment-skill installation instructions. Those behaviors extend beyond the stated monitoring and alerting purpose. The supplied chunk contains no vision, pose-estimation, absence-detection, threshold, or alerting implementation, although its role as a shared utility means their absence here does not establish that the complete skill lacks them.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The supplied code implements an asynchronous video-analysis/reporting client. Its repeated polling checks whether an analysis result is ready; it does not monitor personnel status or trigger warnings. Generic parameters and delegated API methods could support personnel monitoring in an unseen backend, but this chunk does not establish that connection or implement the description's defining real-time detection and alerting behavior. File reading and API submission are supporting implementation details, not independently suspicious capabilities.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
86% confidence
Finding

The visible implementation is a general video-analysis and history-retrieval CLI, materially broader and less specialized than the declared real-time on-duty monitoring skill. It does not visibly implement the defining monitoring, threshold, and warning behaviors. The imported skill implementation is not supplied, so specialized analysis could occur downstream; nevertheless, this chunk does not substantiate the advertised real-time workflow. File output and API delegation are supporting implementation details, not independently suspicious behavior.

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: "staff-absence-detection-analysis"
description: "Real-time monitoring of personnel on-duty status in specific areas based on computer vision and human pose estimation, automatically detects abnormal statuses such as leaving posts and absent from work, supports custom threshold settings, and triggers early warning immediately when abnormality is detected. | 人员离岗实时监测技能,基于计算机视觉与人体姿态估计算法,实时监测特定区域内人员的在岗状态,自动判断离岗、缺岗等异常状态,支持自定义判定阈值,异常发生立即触发预警,适用于工厂车间、监控室

The HTTP helper automatically forwards service credentials to arbitrary absolute URLs.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

Absolute HTTP and HTTPS URLs bypass the configured base URL, but the helper still attaches service tokens and the API key. A caller able to influence the destination can send these credentials to an attacker-controlled host, and plain HTTP additionally permits interception; compromised credentials could expose monitoring data or service access.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The analysis workflow lacks a privacy warning about sending surveillance media to the API service.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

Although the documentation mentions an API, the automatic workflow does not clearly warn that local surveillance media will leave the user's environment or require authorization for that disclosure. Employee imagery and workplace details make unintended third-party processing particularly sensitive.

Content

No source excerpt is available for this finding.

Generic history-report phrases automatically activate this specialized monitoring skill.

Medium
Category
Trigger Abuse
Confidence
94% confidence
Finding

Generic phrases such as 'historical reports' automatically trigger a specialized cloud-history query without requiring personnel-monitoring context. An unrelated request could therefore cause unintended retrieval and display of sensitive employee surveillance reports associated with the automatically selected identity.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to execute analysis, but the program runs analysis even when this flag is supplied.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser accepts --export-env-only and promises not to analyze, but no control-flow branch honors that flag. With media supplied, a caller expecting configuration-only behavior instead initiates analysis, potentially uploading sensitive surveillance footage and incurring service charges.

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 = result_json_common_ai_response

        result_json_health_ai_response = result_json.get("healthAiResponse") if isinstance(result_json,
                                                                                           dict) else result_json
        if result_json_health_ai_response:
            result_json = result_json_health_ai_response

Local video contents are submitted to the analysis API without disclosure.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The function reads the complete local media file and passes its contents to the analysis service without an upload-authorization check or disclosure at this boundary. This is expected analysis functionality rather than demonstrated malicious exfiltration, but callers can inadvertently disclose identifiable employee footage to an external processor.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to execute analysis, but the program still analyzes supplied videos.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only flag is parsed but never checked, so supplied videos still reach analyze_video despite the explicit promise not to execute analysis. This breaks a caller's no-analysis expectation and can cause unintended external media processing or charges.

Content

No source excerpt is available for this finding.

The save() docstring promises to create a new record, but an insertion failure can instead overwrite an existing record.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

Any insertion exception triggers an update by ID and potentially a fallback matching either username or realname, copying fields without verifying ownership or the reason for failure. Conflicting input can therefore overwrite an existing user's identity or credential fields instead of failing creation; practical exploitation depends on who controls incoming models.

Content

No source excerpt is available for this finding.

FileUtil.open silently truncates existing files without disclosing its destructive behavior.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The generically named open helper always uses mode 'w', which immediately erases an existing file's contents. No docstring, explanatory comment, confirmation, or action log discloses that behavior; exception tracing does not warn about successful truncation.

Content

No source excerpt is available for this finding.

Monitoring requests can silently register remote accounts and persist user credentials.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

When credentials are missing, an ordinary HTTP operation can silently request remote registration and persist returned authentication tokens in the local database. This introduces undisclosed identity transmission and credential storage into a personnel-monitoring workflow; the excerpt shows no explicit enrollment consent or dedicated protection for stored secrets.

Content

No source excerpt is available for this finding.

Ordinary HTTP requests can trigger undisclosed remote account registration and identity transmission.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The helper submits the current identity as both openId and mobile with registration enabled, even though its public purpose is making an HTTP request. This hidden side effect can transmit identifying information and create a remote account without an explicit enrollment decision, which is particularly sensitive in a workplace-monitoring context.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
87% confidence
Finding

The POST transmits the current username-derived identity to the configured remote login service and requests registration. External networking alone is not a vulnerability, but this transmission is an automatic enrollment side effect with no explicit consent gate in the supplied code; there is no evidence that the destination itself is malicious.

Content

Scanner excerpt · skills/smyx_common/scripts/util.py (reported line 561)May include surrounding context.

python
"source": ConstantEnum.DEFAULT__SKILL_HUB_NAME
            }
            try:
                _response = requests.post(_url, json=_data)
                if _response.status_code == 200:
                    _response_json = _response.json()
                    if _response_json and _response_json.get("success"):

Loading configuration silently creates directories and a YAML file, including during module import.

Low
Category
Output Handling
Confidence
94% confidence
Finding

YamlUtil.load creates directories and writes a configuration file when the requested path does not exist, without a warning, log, or explanatory docstring or comment. BaseEnum.init_subclass invokes this loader during class definition (L044–L059), so merely importing the module can modify the filesystem without disclosure.

Content

No source excerpt is available for this finding.

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 243)May include surrounding context.

python
username = getattr(model, "username", None)
            if username:
                column_names = self.__model__.__table__.columns.keys()
                update_data = {key: getattr(model, key) for key in column_names if key != "username"}
                return self.update_by_username(username, **update_data)

            return None

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 353)May include surrounding context.

python
column_names = self.__model__.__table__.columns.keys()

            for key in column_names:
                value = getattr(model, key)
                setattr(instance, key, value)

            session.commit()

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 326)May include surrounding context.

python
if filters:
                for key, value in filters.items():
                    query = query.filter(getattr(self.__model__, key) == value)

            if offset:
                query = query.offset(offset)

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 441)May include surrounding context.

python
if filters:
                for key, value in filters.items():
                    query = query.filter(getattr(self.__model__, key) == value)

            return query.scalar()
        finally:

ai_chat documents an agent invocation and analysis result, but its implementation never invokes the agent.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 claims this method invokes openclaw agent through subprocess, and L33–L34 promise an analysis result or session ID. However, command construction and subprocess execution are commented out; L68 creates an empty dictionary, and L70 raises an AttributeError that L81–L82 catch, causing the method to return None rather than perform the documented analysis.

Content

No source excerpt is available for this finding.