Back to skill

Security audit

Reptile Tail Loss (Autotomy) Detection | 守宫/蜥蜴尾巴断尾识别

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report account and credential handling, remote media uploads without clear disclosure, and behavior that does not match its reptile-monitoring purpose.

Consider avoiding installation until the account and credential handling, remote media upload disclosure, and export-only behavior are clarified. Skillspector reports partial analysis because one local path-like reference was missing, so the evidence is incomplete even though all listed components were scanned.

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 (27)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk manages user accounts rather than reptile monitoring. Its concrete model and specialized queries concern user identities, personal data, and tokens; it contains no camera capture, image analysis, tail measurements, injury detection, or tail-loss event recording. Generic database persistence could support the declared system, but the user-specific functionality shown is not explained by that purpose. These operations occur when the DAO is instantiated or its methods are called; the chunk does not demonstrate network transmission or automatic credential collection.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Generic file, date, polling, and HTTP helpers could support the declared monitoring function and are not independently grounds for a mismatch. However, the supplied code also implements an undeclared account-registration, identity-persistence, authentication, and billing workflow. It contains no camera capture, tail-image analysis, historical length comparison, injury detection, or tail-loss event recording. These findings concern capabilities exposed by this code chunk when invoked; they do not establish that every operation runs on import or that the broader skill lacks monitoring functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied chunk materially differs from the declared purpose: it provides generic scaffolding and a disabled chat integration, not a working reptile-tail monitoring system. The commented command does not execute, and the dictionary attribute access prevents the wrapper from producing an analysis. Imported base classes could contain additional functionality, but their implementations are not supplied, so the declared capabilities cannot be attributed to this chunk. No camera access, external command execution, or other hidden capability is established by the visible code.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code implements a video-analysis CLI and history-listing interface. It contains no visible camera capture, scheduling, reptile-tail measurements, historical length comparisons, wound or scab detection, shortening threshold, or timestamped tail-loss alerts. The actual analysis is delegated to an unavailable module, so its domain-specific behavior cannot be verified. Although video processing could support the declared workflow, this chunk exposes a materially different interface and an undeclared history-retrieval capability without demonstrating the claimed monitoring behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied code is a generic API-backed media-analysis wrapper. Its visible behavior does not establish the specialized reptile-tail monitoring workflow described. File upload, polling, and report export could support that workflow, but they do not demonstrate it. Imported API implementations might provide specialized analysis, so the conclusion is limited to the supplied chunk; no malicious behavior is established.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The supplied chunk exposes API-management and user-directory capabilities rather than implementing the declared monitoring behavior. HTTP helpers and download URLs could support image processing and are not inherently mismatches, but the user-lookup capability has no evident relationship to the description. No camera scheduling, visual analysis, historical comparison, injury detection, or event timestamping appears here. Because this is only a shared utility module, the finding is limited to the supplied chunk; it does not establish what the rest of the skill implements.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
80% confidence
Finding

Generic AI-analysis endpoint configuration could support the declared visual-analysis workflow and is not inherently a mismatch. However, the explicit health-service report-export resource is inconsistent with the stated purpose. This chunk only defines configuration; it does not demonstrate actual network requests or data export. The absence of tail-analysis logic in this configuration file alone is not grounds for the 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-reptile-tail-loss-detection-analysis"
description: "Through fixed enclosure cameras, the system periodically captures tail images of geckos and lizards and uses AI visual analysis to detect tail length (compared with historical images or body-length reference values), tail-tip wounds, scabs, or abnormal shortening. | 通过爬宠箱固定摄像头,定期拍摄守宫、蜥蜴等爬行动物的尾部图像,利用 AI 视觉分析技术检测尾巴长度(与历史图像或同体长参考值对比)、尾部尖端伤口、结痂或异常短缩。当检测到尾巴长度突然明显缩短(例如缩短超过 20%)、尾部断�

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 local-media analysis workflow lacks a clear warning that input images or videos are sent to a remote API.

Medium
Category
Output Handling
Confidence
89% confidence
Finding

The local-file workflow does not clearly state that image or video contents leave the device for remote processing. General references to APIs and cloud history are not an explicit upload notice, and enclosure footage can incidentally reveal people, private interiors, or breeding-facility information.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to run analysis, but does not prevent analysis execution.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The export-only flag is parsed but never checked, so supplying it with media input still invokes analysis. A user relying on the explicit no-analysis promise can unintentionally trigger remote processing, media transfer, or service charges; identity initialization also occurs before any mode-specific handling.

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 file contents are submitted to the analysis API without disclosure that the file will be uploaded.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The local-file branch reads the entire media file and passes its bytes and filename to the analysis API without a visible disclosure or authorization gate. Users expecting local processing could consequently expose enclosure footage and incidental private information to a remote service.

Content

No source excerpt is available for this finding.

The documented export-only mode still executes analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The documented export-only option is never consulted after parsing, allowing execution to continue into analysis when input is supplied. This violates an explicit no-analysis expectation and can cause unintended remote processing or charges, while identity initialization occurs even before the input checks.

Content

No source excerpt is available for this finding.

Database initialization modifies a user-account schema unrelated to reptile tail-loss detection.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

Every Dao initialization invokes _alter_tables() (L185-L186), which inspects sys_user and adds source_id and realname columns when absent (L200-L215). This is account-management functionality, supported by a User model containing personal information and authentication-token fields (L448-L465), rather than storage of reptile images, measurements, or tail-loss events. The stated purpose does not justify automatically modifying this unrelated user-account schema.

Content

No source excerpt is available for this finding.

ai_chat documents an OpenClaw agent invocation, but its execution path never invokes an agent or produces analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The execution is disabled, and accessing stderr on an empty dictionary always raises an exception that is swallowed, leaving callers with None instead of analysis or an explicit failure. If this helper supports tail-loss monitoring, silent failure can suppress detection or alerts, although the supplied source does not establish that the monitoring workflow calls it.

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
97% confidence
Finding

Opening a path in 'w' mode immediately erases any existing contents, even before the caller writes data. This generic open helper has no docstring, comment, confirmation, or operation-specific log explaining that behavior; its exception handler does not disclose successful truncation.

Content

No source excerpt is available for this finding.

The request helper silently registers service accounts and persists account credentials, beyond the declared tail-analysis workflow.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

An ordinary HTTP request can initiate silent remote registration using the current identity and persist returned bearer credentials in SQLite without an explicit enrollment step. This introduces identity disclosure and credential-storage side effects not evident from the tail-analysis description; the shared-library context explains the implementation but does not provide consent or storage protection.

Content

No source excerpt is available for this finding.

HTTP requests can silently trigger remote account registration and local credential persistence.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The generic request helper hides remote account creation and local credential persistence behind routine request handling. This is the same underlying enrollment issue as SDI-1, not a separate exploit, and can create persistent service identities without a deliberate user decision.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The POST transmits the selected identity as both openId and mobile to a configured registration endpoint without explicit enrollment consent. The caller prioritizes API_SECRET_KEY as that identity, so a configured secret can also be copied into registration fields; the source does not establish that the endpoint is attacker-controlled.

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 writes missing configuration files.

Low
Category
Output Handling
Confidence
94% confidence
Finding

YamlUtil.load creates parent directories and writes a YAML file when the requested path does not exist, without a disclosure comment, docstring, prompt, or log. This unexpected write can also occur during module import through BaseEnum.init_subclass (L044–L059), rather than through an explicit save operation.

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)