Back to skill

Security audit

Smart E-Bike Detection Skill | 电动车智能检测技能

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because it reports hidden account and credential handling, authenticated requests to caller supplied URLs, and behavior that does not consistently match the stated vehicle detection purpose.

Consider avoiding installation unless you need the described vehicle detection and are comfortable with the reported remote account, credential, and media handling. The scan is partial: it reports all 28 listed components were scanned, but one reference was unresolved, so the evidence is incomplete. The findings support a suspicious review verdict; they do not establish malicious intent or that the reported behaviors have been invoked.

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/electric_vehicle_detection_analysis.py:211
Finding

Upstream sender identity is replaced with a workspace-shared account

Content
View full analysis

Impact

A second sender can retrieve reports accumulated under the shared default account, including reports created by another sender. New analyses are also attributed to that shared account.

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 (25)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

Ordinary file, JSON, polling, and HTTP helpers can support a computer-vision skill and are not inherently mismatches. However, this chunk also implements an undisclosed identity/account lifecycle: reading a local identity file, generating persistent users, silently registering or logging them into a remote service, and persisting tokens. The HTTP wrapper can additionally attach configured credentials to caller-supplied destinations without a visible host restriction. These behaviors extend beyond the declared detection and alerting purpose. No vision processing, violation counting, or alert generation appears in this chunk, although its utility-module role means that absence alone does not establish that the complete skill lacks those features. The supplied code does not demonstrate that these callable behaviors have actually been invoked.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The visible implementation does not substantiate the declared computer-vision functionality and instead consists of shared scaffolding and a nonfunctional agent-chat stub. It does not actually execute the commented-out external command. Imported base classes and API services are not supplied, so they could contain additional functionality; this assessment is limited to the provided chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Generic database persistence could support the declared skill, but this chunk specifically implements user identity and token management rather than detection results, violation counts, or alerts. That concrete capability is not represented in the description. No computer-vision, image/video processing, or alerting behavior appears in the supplied chunk. This assessment is limited to this file; it does not establish what the rest of the skill implements or indicate malicious 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: "electric-vehicle-detection-analysis"
description: "Automatically detects electric motorcycles and e-bikes in restricted areas based on computer vision. It supports real-time detection for both video streams and images, counts the number of illegal parking or driving instances, and triggers violation alerts to assist with safety management in parks, communities, and organizations. | 电动车智能检测技能,基于计算机视觉自动检测禁行区域内的电动摩托车/电动车,支持视频流和图片实时检测,统计违规停放/行驶数量,触发违规预警,助力园区/社区/单位安全管理"
version:

Caller-supplied external URLs receive the skill's configured credentials and user identity.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The wrapper accepts arbitrary absolute HTTP or HTTPS URLs but automatically attaches configured access tokens, API keys, authorization credentials, and user identity. A caller-controlled destination can therefore receive service credentials, and accepting plain HTTP also permits interception in transit.

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 --export-env-only option promises not to execute analysis, but analysis still runs.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

The parser accepts --export-env-only and promises not to analyze, but execution never checks the flag before initializing identity and invoking analysis or report queries. A user relying on this no-analysis mode can unintentionally submit sensitive surveillance media or initiate remote operations.

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 API without disclosure of the upload.

Medium
Category
Output Handling
Confidence
87% confidence
Finding

The function reads the entire local file, packages its contents as an upload, and passes it to the analysis API. Although a comment explains the file read, this file contains no prompt, log, docstring, or comment disclosing that the video is transmitted to an external service; requesting video analysis does not necessarily imply consent to remote processing.

Content

No source excerpt is available for this finding.

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

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only flag is parsed but never honored, so supplying media still reaches analysis and --list still reaches report retrieval. This violates the advertised non-analysis boundary and can cause unintended transmission or processing of sensitive video.

Content

No source excerpt is available for this finding.

The skill includes user-profile and credential database management unrelated to electric-vehicle detection.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

L448-L465 define persistent user records containing personal information and authentication tokens, and L470-L473 expose those records through inherited CRUD operations. Managing user identities and credentials is not justified by the manifest's stated image/video detection, violation counting, and alerting purpose; this is distinct from storing detection results.

Content

No source excerpt is available for this finding.

The ai_chat docstring claims to invoke the OpenClaw agent and return analysis results, but the implementation never invokes it and returns no result.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26–L34 explicitly promises a subprocess-based agent call and an analysis result or session ID. However, the command and subprocess execution are commented out (L48–L66), and the empty dictionary assigned at L68 causes result.stderr at L70 to raise an exception that is caught at L81–L82, leaving the method to return None. This is an operational contradiction of the documented analysis capability, not evidence of actual subprocess execution.

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
93% confidence
Finding

Opening the path in 'w' mode immediately erases existing contents, even before the caller writes anything. The generic open method has no docstring, comment, confirmation, or action log disclosing this behavior; exception tracing does not disclose successful truncation.

Content

No source excerpt is available for this finding.

The detection request wrapper silently provisions remote accounts and persists their credentials.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

An ordinary HTTP operation can trigger remote login or registration using the current identity and then persist returned credentials, without an explicit provisioning decision in this interface. For a vehicle-detection skill, this is a consequential hidden account-management side effect that can disclose identifiers and create unwanted persistent accounts.

Content

No source excerpt is available for this finding.

The HTTP helper silently performs remote login or registration and persists returned credentials.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The generic request helper implicitly performs registration or login and saves returned authentication material when tokens are missing. These account-changing and persistent side effects are not necessary properties of an HTTP helper and can occur without a distinct user-approved setup step.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
86% confidence
Finding

This POST transmits an identity to a separate login service as both openId and mobile while enabling registration. The caller can select API_SECRET_KEY as that identity, so a configured API secret can also be copied into account-identification fields; the actual destination and its trust relationship cannot be verified from this file.

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"):

The mandatory report format fixes user-visible labels in Chinese without a language choice.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The fixed output-column specification mandates Chinese column labels, a Chinese report-name template, and Chinese link text. No user language choice or region-specific justification is provided for imposing that output language.

Content

No source excerpt is available for this finding.

Generic history-report phrases automatically trigger an e-bike report query without a domain check.

Low
Category
Trigger Abuse
Confidence
93% confidence
Finding

Generic requests such as viewing historical reports are instructed to trigger a mandatory remote e-bike report query without checking the intended domain. In an agent with multiple skills, this can retrieve and display surveillance-related records outside the user's intended task.

Content

No source excerpt is available for this finding.

Configuration loading silently creates directories and files, including during module import.

Low
Category
Output Handling
Confidence
91% confidence
Finding

YamlUtil.load creates parent directories and writes a configuration file when the requested path is absent, without a comment, docstring, prompt, or log disclosing that side effect. BaseEnum.init_subclass invokes this loader when the enum classes are defined (L044–L059), so merely importing the module can modify the filesystem.

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:

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/util.py (reported line 30)May include surrounding context.

python
elif isinstance(cur, dict):
                        cur = cur.get(part, default)
                    else:
                        cur = getattr(cur, part)
                return cur
            except Exception:
                return default