Back to skill

Security audit

Commuter After-Work Fatigue Care (Home-Arrival Moment) | 上班族下班疲劳关怀(回家时刻)

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with high confidence because scanners report privacy and account-handling concerns, unexpected behavior for its stated home fatigue-care purpose, and a data-flow issue.

Treat this as a review warning before installing. Consider whether you are comfortable with external processing of home video and with the reported account registration, identity and token handling; the scanners say coverage was partial because a local reference was missing, so the evidence is incomplete.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/smyx_commuter_fatigue_care_analysis.py
Finding

CLI identity initialization discards the upstream sender and reuses a workspace-wide account

Content
View full analysis

Impact

A sender can receive reports created through another sender's normal Skill invocations under the same fallback account. New analyses are also attributed to that shared account. This exposes potentially sensitive home and fatigue reports in shared-Agent deployments.

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
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (28)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk does not implement or visibly invoke the declared monitoring and intervention workflow. The only substantive local method concerns an agent-chat interface, and even that cannot execute its intended command because subprocess execution is disabled. Imported base/API classes could contain additional behavior, but their implementations are not supplied and cannot establish that the declared purpose is fulfilled. No active subprocess execution or other hidden capability should be inferred from commented-out code.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Generic helper functions alone would not establish a mismatch. However, this chunk includes a substantive account-registration, credential-management, and remote API layer that is not explained by the declared camera-based fatigue-monitoring purpose. It contains no camera or audio acquisition, posture or facial analysis, sigh counting, 30-minute monitoring window, fatigue scoring, or smart-speaker control. The assessment is limited to the supplied chunk; it does not establish that these functions are invoked elsewhere or that malicious exfiltration occurs.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The code's visible functionality centers on submitting supplied videos and managing API-generated analysis reports. Video submission could support the declared analysis, and an external service might perform fatigue detection, but the supplied chunk neither specifies those detectors nor implements the timing and smart-speaker actions central to the description. Therefore, the declared end-to-end behavior is materially different from the behavior demonstrated here. The imported API implementations are not supplied, so their internal analysis capabilities cannot be verified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied chunk is a user-database component rather than a fatigue-monitoring implementation. Database persistence could be a supporting implementation detail, but this code specifically manages identity, profile, and token data without a visible connection to fatigue observations or smart-home responses. It contains no camera/audio processing, 30-minute observation logic, fatigue scoring, speaker announcements, or music control. This assessment is limited to the supplied chunk; it does not establish that other files lack those features or that stored tokens are exfiltrated.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The supplied code implements a general video-analysis CLI and an additional history-listing operation. It contains no visible camera monitoring, arrival detection, 30-minute tracking, fatigue scoring, or smart-speaker speech/music control. Analysis is delegated to an unavailable implementation, so the specific fatigue detectors cannot be confirmed or ruled out. Nevertheless, the visible workflow differs materially from the description, and user-associated analysis-history retrieval is an undeclared capability. Identity resolution and optional output-file writing are supporting implementation details, not independently treated as mismatches.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

Local and remote video inputs, report formatting, and file output are reasonable supporting details. The material extension is a separate, keyword-triggered history-retrieval function absent from the declaration. The snippet does not establish whether fatigue detection or smart-speaker actions are implemented, so their absence from these command examples is not itself treated as a mismatch.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

Generic AI-analysis configuration could support the declared purpose and is not inherently a mismatch. However, the explicit health-order report export resource introduces an unrelated capability not explained by that purpose. This finding is limited to endpoint configuration: the chunk does not demonstrate actual report access or export. The absence of camera, fatigue-detection, and speaker logic in this configuration file is not itself treated as a mismatch.

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-commuter-fatigue-care-analysis"
description: "Through a fixed camera in a smart-home living room, the system analyzes office worker behavior in the first 30 minutes after coming home, detecting slumped sitting / reclining (relaxed posture, back-to-sofa angle > 120°), facial fatigue features (visible eye bags, downturned mouth corners, frequent blinking), and sighing frequency (rapid chest/abdomen rise-fall with audible exhale). | 通过智能家居客厅的固定摄像头,分析上班族回家后30分钟内的行为,检测瘫坐/斜躺(姿态放松、背部与沙发夹角>120°)、面部疲惫特征(眼袋明显�

The analysis workflow sends private home-camera inputs to an API without a clear external-processing privacy warning.

High
Category
Output Handling
Confidence
95% confidence
Finding

The workflow directs API-based processing of private living-room footage without clearly obtaining informed consent for transferring raw media to an external processor. References to cloud reports are not an explicit raw-video upload disclosure, and the prohibition on third-party sharing can give users conflicting privacy expectations.

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 documented export-only mode still executes video analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser accepts --export-env-only and promises no analysis, but main never checks the flag before initializing identity and invoking analysis or history retrieval. With a video input, a user expecting configuration-only behavior can instead initiate external processing of sensitive home footage.

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 an analysis API without an upload or privacy disclosure.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The code reads the complete local video into a multipart file payload and passes it to the analysis API without an upload disclosure or consent check in this path. For living-room footage, this crosses a sensitive privacy boundary and may expose household members who did not expect external processing.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to execute analysis, but the code still invokes video analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only argument is parsed but never used to alter control flow, so supplying input still reaches analyze_video despite the explicit promise not to analyze. In this skill's context, that can cause unexpected external processing of private video; identity initialization and history requests are also not suppressed.

Content

No source excerpt is available for this finding.

L098–L101 expose username-based user lookup outside the declared fatigue-care workflow.

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

The get_user_by_username method delegates username-based user retrieval to RequestUtil. The manifest describes camera-based fatigue analysis and smart-speaker responses, neither of which establishes a need for user-directory lookup.

Content

No source excerpt is available for this finding.

The skill includes persistent identity and authentication-token management unrelated to fatigue monitoring.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The User model stores real names, email addresses, birthdays, and authentication tokens (token and open_token), and UserDao exposes inherited CRUD operations for these records. Managing persistent identity and credential records is not justified by the stated camera-based fatigue analysis and smart-speaker care workflow; database initialization also creates and upgrades this user table (L180–L215).

Content

No source excerpt is available for this finding.

ai_chat documents invoking an analysis agent, but its implementation never invokes one and silently returns None.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 explicitly states that this method invokes openclaw agent through subprocess, and L33-L34 promise an analysis result or session ID. However, the invocation is commented out at L61-L66; L68 instead creates a dictionary, whose nonexistent stderr attribute raises an exception that L81-L82 suppress, leaving callers with None rather than an analysis result.

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
94% confidence
Finding

Opening a path in 'w' mode immediately erases any existing contents, even before the caller writes data. This generically named helper has no docstring, comment, confirmation, or operation log explaining that behavior; its exception handler only reports failures.

Content

No source excerpt is available for this finding.

The HTTP helper silently registers or logs into a remote health-service account and persists account credentials, beyond the declared fatigue-analysis and speaker-control purpose.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

A generic HTTP operation can silently register or log into a remote account, transmit an identity, and persist returned bearer credentials without a separate enrollment decision. In a home-monitoring skill, this introduces an undisclosed identity-linkage and credential-retention boundary beyond merely analyzing fatigue.

Content

No source excerpt is available for this finding.

Ordinary HTTP requests can silently transmit an identity and register a remote account.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The request helper performs a state-changing account operation using register=1 and silent=1 as an implicit authentication fallback. An ordinary service request can therefore disclose an identifier and create an external account without a distinct authorization step, which is particularly sensitive for household monitoring.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

This POST sends the selected identifier as both openId and mobile to the configured health service without a visible consent gate or HTTPS requirement. The caller preferentially selects API_SECRET_KEY as that identifier, so a configured API secret can also be copied into identity fields and disclosed outside its intended authentication channel.

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"):

L017–L020 silently create directories and a configuration file during a load operation, including when invoked during module import.

Low
Category
Output Handling
Confidence
90% confidence
Finding

YamlUtil.load writes a new YAML file when the requested path is absent, without a warning, log, docstring, or explanatory comment disclosing this side effect. BaseEnum.init_subclass invokes this loader during class creation (L052–L058), so importing the module can modify the filesystem without an explicit save request.

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 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)