Back to skill

Security audit

Hydroponic Nutrient Concentration Visual Assessment | 水培植物营养液浓度视觉评估

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report account and credential handling, media uploads, and behaviors that diverge from its stated hydroponic plant-analysis purpose.

Consider not installing until the account and credential handling, external media uploads, and export-only behavior are clearly explained and verified. The scan is partial: two referenced local paths were missing, so the evidence does not establish what those parts contain.

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 contains no camera capture, image processing, AI vision, plant-health classification, nutrient assessment, or adjustment recommendations. Although database storage could support a plant-monitoring application, this implementation specifically manages users, personal information, and tokens, with no demonstrated connection to the declared hydroponic workflow. Its primary behavior and data resources therefore materially differ from the description. The chunk does not show token exfiltration or external transmission.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk does not implement the declared workflow, and its only explicit analysis-related method is nonfunctional. Although comments describe invoking an external agent, that execution is disabled, so it should not be treated as an active capability. Inherited classes may provide additional functionality, but their implementations are not supplied; this assessment is limited to 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 presents a general video-analysis interface and a separate history-listing operation. It contains no visible plant-feature assessment, nutrient-concentration logic, or adjustment recommendations. Those domain-specific functions could exist inside the imported skill implementation, which is not supplied, so their absence here alone does not establish that they are unavailable. However, the visible generic video workflow and user-associated history capability materially diverge from the stated specialized plant-image workflow.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Generic formatting, polling, and HTTP helpers could legitimately support the declared skill and are not themselves a mismatch. However, the supplied chunk adds substantive identity provisioning, credential-file access, remote account authentication, and billing behavior that the description does not disclose. It contains no camera capture, plant-image analysis, nutrient assessment, or adjustment logic. Because this is a shared utility module, that absence alone does not establish what the full skill does; the mismatch rests on the additional account and credential operations actually implemented here. No malicious intent or confirmed data exfiltration is established.

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 utility rather than the described hydroponic vision workflow. Networking and download helpers could support that workflow, so their presence alone would not establish a mismatch. However, the exposed user-lookup and general data-management capabilities have no demonstrated relationship to the declared purpose. These are method definitions; the chunk does not establish that any requests are actually executed or reveal their destinations. Other files could implement the declared workflow, but that cannot be verified from this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding

Generic AI-analysis endpoints could support the declared purpose and are not inherently mismatched. However, DETAIL_EXPORT_URL explicitly targets '/health/order/api/getReportDetailExport?id=', introducing a health-report resource not explained by the description. This chunk only configures URLs; it does not demonstrate an actual request or export. The absence of plant-analysis logic in a configuration file alone is not evidence of a mismatch.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding

The visible code is primarily an analysis-submission and report-management wrapper. It contains no fixed-camera capture, root-color or leaf-morphology assessment, nutrient-concentration determination, or adjustment-recommendation logic. Those functions could exist in the imported service or remote backend, so their absence here does not establish that they are absent from the complete system. However, the supplied chunk exposes a broader generic video/file-analysis interface and historical-report retrieval that the plant-specific description does not accurately convey. File reading, API submission, and polling are supporting implementation details, not independently suspicious capabilities.

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-hydroponic-nutrient-assessment-analysis"
description: "Using fixed cameras on a hydroponic growing rack to capture high-resolution images of plant roots (in transparent containers) and leaves (young and old), AI vision analysis identifies root color (white = healthy, yellow = early stress, brown = severe stress, black = rotting) and leaf morphology (tip burn, leaf-margin scorch, yellowing, curling) to judge whether the nutrient solution is too concentrated or too dilute, and. | 通过水培种植架的固定摄像头拍摄植物根系(透明容器)和叶片(新叶、老叶)的高清图像,利用AI视觉分析技术识别根

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 local images or videos to an external API.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The workflow directs the agent to process local media through an external API without clearly warning that the media itself leaves the device. Although plant imagery is generally less sensitive than personal media, growing-facility images can expose people, locations, proprietary operations, or embedded metadata.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to run analysis, but the program ignores it and executes analysis normally.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The program accepts --export-env-only but never checks it, despite promising not to execute analysis. When media input is also supplied, a caller relying on that promise can trigger normal analysis and external data transfer; identity initialization also occurs before any such guard.

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 passed to an analysis API without disclosure of the upload.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The implementation reads the complete local media file and passes its bytes to the analysis API without an upload disclosure or consent check in this path. Plant media can still contain sensitive facility details, people, or metadata, so selecting a local file should not silently imply permission for external transfer.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to execute analysis, but the code ignores it and follows the normal analysis path.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only flag is parsed but never used to stop execution. With media input present, users expecting a non-executing configuration operation instead trigger normal analysis, potentially uploading private media and consuming service resources.

Content

No source excerpt is available for this finding.

The skill includes user-profile and credential database management unrelated to hydroponic nutrient assessment.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

L448-L473 define a persistent user model containing personal information and token/open_token fields, and expose it through UserDao with inherited create, read, update, and delete capabilities. Database initialization also creates and migrates the sys_user table (L184-L215). Managing user identities and credentials is not justified by the manifest's stated purpose of analyzing plant images and recommending nutrient adjustments.

Content

No source excerpt is available for this finding.

The AI-analysis helper claims to invoke an agent and return results, but its implementation never executes the agent.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26–L34 explicitly promises a subprocess call to the OpenClaw agent and an analysis result or session ID. However, execution is commented out at L61–L66, and accessing stderr on the empty dictionary at L68–L70 raises an exception that is caught without returning a result. This contradicts the documented analysis behavior rather than merely changing an implementation detail.

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

At L101, the generic open helper uses mode 'w', which immediately truncates an existing file. No comment, docstring, confirmation, or user-facing message explains this behavior; exception tracing does not disclose successful overwrites.

Content

No source excerpt is available for this finding.

The request helper silently provisions service accounts and persists user credentials, beyond the stated plant-image assessment purpose.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

An ordinary request can trigger silent remote registration and persist returned profile data and authentication tokens in a local SQLite database without dedicated storage protections shown. The hydroponic purpose does not make this implicit identity lifecycle apparent, creating unexpected privacy and credential exposure.

Content

No source excerpt is available for this finding.

Ordinary HTTP requests can silently register a remote account using the current identity.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The request helper sends silent=1 and register=1 to a remote login endpoint automatically when authentication tokens are unavailable. This turns routine plant-analysis requests into undisclosed account creation using the current identity, crossing a privacy and authorization boundary.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The login request transmits the supplied username as both openId and mobile, while the surrounding code preferentially selects API_SECRET_KEY as that username. Consequently, a configured secret can be disclosed as identity data to the separately configured health endpoint, potentially over HTTP, rather than remaining 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"):

The service exposes username-based user lookup, which is unrelated to hydroponic nutrient assessment.

Low
Category
Excessive Agency
Confidence
92% confidence
Finding

At L098–L101, get_user_by_username forwards a supplied username to a user-lookup utility. Looking up user accounts is not a direct requirement of analyzing root and leaf images or recommending nutrient adjustments, and the manifest does not declare account-management functionality.

Content

No source excerpt is available for this finding.

Configuration loading creates directories and files without disclosure.

Low
Category
Output Handling
Confidence
93% confidence
Finding

YamlUtil.load silently creates a directory and writes a YAML file when the requested configuration is absent, with no explanatory comment, docstring, or user-facing message. Because BaseEnum.init_subclass invokes this loader during class creation (L052–L058), importing the module can modify the filesystem despite appearing to only load configuration.

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()