Back to skill

Security audit

Reptile Circadian Activity Analysis | 爬宠活动量昼夜节律分析

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners report undeclared account, credential, and remote data-handling capabilities that do not fit its advertised reptile activity monitoring purpose.

Treat this as a review-needed skill and consider avoiding installation until its publisher explains why account creation, identity and token storage, and remote video analysis are needed, and clarifies what the export-only option actually does. Scanner coverage is marked partial because a referenced local path was missing, so the evidence does not establish that all artifact behavior was assessed.

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

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk implements account, authentication, and API-client infrastructure rather than camera capture, motion analysis, hourly aggregation, circadian charting, or multi-day anomaly detection. Generic file, date, and polling helpers could support the declared purpose, but the identity-file access, persistent user management, remote registration/login, and payment workflow introduce material capabilities not disclosed by it. These operations are primarily exposed through callable methods; the chunk does not show that they run automatically or establish malicious exfiltration. The assessment is limited to the supplied code.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The visible implementation is materially different from the declared monitoring workflow. Although it contains comments about invoking an external agent, that execution is disabled and does not establish an active subprocess capability. Imported base classes and services are not supplied, so their behavior cannot be assessed; the mismatch finding is limited to the visible code.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Database persistence could support the declared purpose, but this chunk specifically implements user and credential-record management rather than storage of video, motion measurements, or circadian results. Initialization creates the registered schema and migrates the sys_user table. No camera capture, motion detection, hourly aggregation, charting, or rhythm-warning logic appears in the supplied chunk. This establishes a mismatch for the supplied code, without establishing what the rest of the skill does. No network transmission or exfiltration is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk operates on already-existing video inputs and generic API reports. It does not access a camera or maintain the continuous capture and multi-day monitoring workflow central to the description. Imported API implementations might perform some specialized analysis, but their behavior is not shown and cannot establish that the declared functionality exists. Video validation, upload preparation, and polling could reasonably support the declared analysis; the mismatch is the materially different visible workflow and additional report-history functionality, not those supporting details alone.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The code requires --input or --url for analysis and delegates processing to an imported skill object. It also exposes --list to retrieve analysis history after resolving an open ID. No camera acquisition, 24-hour recording, hourly motion aggregation, activity-chart generation, or multi-day abnormality detection is visible. Imported implementations could provide some specialized analysis, but the supplied chunk does not establish those behaviors, and its history-retrieval capability is outside the declared purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk implements a general-purpose API client rather than camera capture, motion measurement, hourly aggregation, charting, or circadian anomaly detection. Network helpers could support a larger monitoring system, but this chunk does not establish that connection, particularly for user lookup and generic data-management operations. These methods expose capabilities; the snippet does not show them being invoked, identify their destination services, or establish malicious activity.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

Generic analysis endpoints could support the declared workflow, so they are not independently evidence of a mismatch. However, the explicit health-order report export endpoint introduces a materially unrelated resource not explained by the description. This chunk contains no camera capture, motion measurement, hourly aggregation, charting, or multi-day rhythm checks, although their absence from a configuration-only file does not establish that they are absent from the full skill. The mismatch finding rests on the unrelated health-report configuration, not on any demonstrated network access or data export.

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-reptile-circadian-activity-analysis"
description: "Through a fixed camera in a reptile enclosure, the system continuously captures 24-hour video and uses motion-detection techniques to count hourly activity volume (pixel-change area or motion-pixel ratio), producing a circadian activity distribution chart. | 通过爬宠箱固定摄像头,连续 24 小时采集视频,利用运动检测技术统计每小时的活动量(像素变化面积或运动像素比例),生成昼夜活动分布图。当节律异常持续多日时,输出'昼夜节律紊乱'提示,建议调整光照周期或检查环境干扰(如夜间灯光、

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 --export-env-only option promises not to perform analysis, but the flag is ignored.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser accepts --export-env-only and promises no analysis, but execution never checks the flag. When supplied with an input, the command still initializes identity and invokes analysis, potentially transmitting private enclosure footage or creating remote records despite the user's explicit non-analysis selection.

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

The skill passes complete enclosure videos to an API service without disclosing video upload in its manifest.

Medium
Category
Excessive Agency
Confidence
87% confidence
Finding

The manifest describes fixed-camera monitoring and motion-pixel aggregation, but L125-L131 read the entire input video into an upload payload and L136-L139 pass it to the API analysis service. Transferring raw enclosure footage is a material data-handling behavior beyond the described activity-counting workflow, rather than an inherent requirement of calculating hourly motion statistics.

Content

No source excerpt is available for this finding.

Local video contents are passed to the analysis API without disclosure of the upload.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The code reads the entire local file and passes its contents to the analysis service through the files argument. The provided file contains no confirmation, user-facing notice, or comment explaining this transmission; requesting video analysis does not necessarily imply consent to upload a local video to a remote service.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to analyze video, but does not prevent analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only flag is parsed but never used to stop execution. With an input or --list, the command proceeds into identity initialization and remote operations, violating the promised non-analysis mode and potentially exposing footage or accessing account data unexpectedly.

Content

No source excerpt is available for this finding.

The service exposes username-based user lookup, which is unrelated to reptile activity analysis.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

At L098–L101, get_user_by_username forwards a supplied username to RequestUtil.get_user_by_username, providing an account-lookup capability. Looking up user accounts is not justified by the manifest's camera capture, motion measurement, circadian charting, or environmental recommendations; this file does not establish that the capability is actually invoked.

Content

No source excerpt is available for this finding.

L448–L473 implement user-profile and authentication-token persistence unrelated to reptile activity analysis.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The model stores reusable authentication tokens as ordinary strings in a workspace-shared SQLite database, without application-level encryption or explicit restrictive file permissions in the shown DAO. Readers of that database or its backups could obtain credentials; authentication support may be legitimate shared infrastructure, but it introduces sensitive storage beyond the advertised reptile-analysis task.

Content

No source excerpt is available for this finding.

The ai_chat docstring promises an OpenClaw agent subprocess call, but the implementation never invokes the agent or returns its analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 explicitly states that this method calls the OpenClaw agent through subprocess. However, the command and subprocess invocation are commented out, and L68–70 instead access stderr on an empty dictionary, raising an exception that is caught at L81–82. Consequently, the method returns None rather than performing the documented analysis.

Content

No source excerpt is available for this finding.

FileUtil.open silently truncates existing files without disclosing its destructive write mode.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

Opening a path with mode 'w' immediately erases existing contents, even if the caller never writes anything. This generically named helper has no docstring, comment, confirmation, or user-facing log explaining that behavior; its exception handling only reports failures.

Content

No source excerpt is available for this finding.

The HTTP helper silently provisions remote user accounts and persists authentication credentials, capabilities unrelated to reptile activity analysis.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

An ordinary HTTP request can implicitly register or log in a remote identity and save returned authentication tokens to a local database. This creates undisclosed external and persistent side effects, and the identity can be sourced from API_SECRET_KEY, exposing more sensitive information than reptile activity analysis requires.

Content

No source excerpt is available for this finding.

HTTP requests can silently trigger remote login or account registration without disclosing the identity transmission.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The helper submits identity fields to /sys/phoneLogin with both silent and register enabled as an automatic consequence of a normal request. Without a distinct authorization step in this path, callers can trigger remote account creation and identity disclosure unexpectedly.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

This POST sends the selected username as both openId and mobile to a configured remote health-service endpoint; the enclosing caller prioritizes API_SECRET_KEY as that username. Consequently, a configured API secret can be disclosed as account-profile data, potentially entering server-side records and logs rather than being confined to 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"):

Loading configuration silently creates directories and files, including during module import.

Low
Category
Output Handling
Confidence
94% confidence
Finding

YamlUtil.load creates a parent directory and writes a YAML file when the requested configuration is absent, without a warning, log, or explanatory comment. BaseEnum.init_subclass invokes this loader during class definition (L044–L059), so merely importing this module can modify the filesystem unexpectedly.

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()