Back to skill

Security audit

Leaf Aging Fall Prediction | 植物叶片老化/脱落预测

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report substantial account, credential, and data-handling behavior that is not clearly aligned with its stated plant-monitoring purpose.

Treat the suspicious verdict as a reason to avoid installing until the account, credential, and video-upload behavior is clearly explained and the reported issues are addressed. The scan was partial: one local reference was missing, so some evidence remains incomplete; it does not prove malicious intent.

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 (29)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk is a user-management persistence module rather than a plant-image analysis component. Database storage alone could be a supporting implementation detail, but the concrete sys_user model and account-specific operations manage names, email addresses, birthdays, demographic fields, and tokens, with no demonstrated connection to the declared workflow. The chunk contains no camera capture, leaf analysis, time-series prediction, reporting, or notification logic. This assessment is limited to the supplied code; it does not establish what other skill components implement or indicate malicious behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied code implements a generic API-backed video-analysis and report-management interface. It contains no plant-specific processing or orchestration establishing the declared monitoring and forecasting purpose. Although the imported API could perform analysis not visible here, its implementation is not supplied, so the claimed leaf-aging and leaf-drop prediction behavior is not supported by this chunk. The mismatch concerns the materially different visible workflow, not the mere use of an API as an implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied chunk contains no camera capture, leaf-image analysis, time-series prediction, daily reporting, or leaf-drop notification logic. Generic utility functions alone would not establish a mismatch, but automatic account provisioning, credential-file access, token persistence, credential-bearing network requests, and payment-related handling introduce substantial behaviors not disclosed by the plant-monitoring description. These behaviors occur when the relevant methods are invoked; this chunk does not establish that they run automatically or that credentials are actually sent to an untrusted destination. Other files could implement the advertised plant features, but that would not resolve the undeclared account and authentication capabilities shown here.

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 an on-demand video-analysis CLI and history-listing interface, materially different from the described continuous plant-aging prediction system. Analysis is delegated to imported code whose behavior is not supplied, so plant-specific functionality inside that dependency cannot be ruled out. Nevertheless, this chunk provides no evidence of the central camera-monitoring, forecasting, or alerting workflow, while explicitly exposing generic video analysis and account-associated history retrieval.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The visible implementation does not substantiate the declared primary purpose. Its only substantive helper is an incomplete generic agent-chat wrapper, not a plant-aging analysis pipeline. No subprocess execution actually occurs because that code is commented out. Imported base classes and services are not supplied, so their behavior cannot be assessed; this finding is limited to the visible chunk and does not establish malicious behavior or undeclared external access.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied chunk implements API-service utilities rather than camera capture, visual analysis, temporal prediction, reporting, or alerts. HTTP utilities could support the declared system, so networking alone is not a mismatch; however, the user-lookup functionality and general remote-management interface are not explained by the declared purpose. This assessment is limited to the supplied chunk: it defines these capabilities but does not itself invoke them, and the destination endpoints and underlying request behavior are not shown.

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 functionality and are not inherently a mismatch. However, the explicit health-service URL ending in /health/order/api/getReportDetailExport?id= points to an unrelated resource not explained by the description. This chunk only defines configuration; it does not demonstrate actual network requests or report exports. Missing plant-analysis logic alone is not conclusive because only a configuration file was supplied.

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-leaf-aging-fall-prediction-analysis"
description: "Using a fixed indoor camera to continuously capture leaf images of houseplants from the same angle every day, AI vision techniques detect leaf color changes (green → yellow → brown), loss of glossiness (reduced surface reflectance), and formation of the abscission zone at the petiole base (angle change). | 通过室内绿植固定摄像头连续采集叶片图像(每天同一角度),利用AI视觉分析技术检测叶片颜色变化(从绿到黄再到褐)、光泽度下降(叶面反光减弱)、叶柄基部离层形成(角度变化)等老化进程,并基�

The request helper attaches service credentials to arbitrary caller-supplied destinations.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

Caller-supplied absolute HTTP or HTTPS URLs are accepted, but service tokens and the API key are attached without checking the destination. If an attacker influences the URL, credentials can be sent to an attacker-controlled host; HTTP also exposes them to network interception.

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 analysis workflow lacks a privacy warning about sending indoor camera imagery to an external API.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

Although cloud API use is mentioned, the workflow does not clearly warn that local indoor-camera media will leave the device or explain the recipient and retention implications. Plant footage can also reveal people, home interiors, and routines, making undisclosed upload a meaningful privacy risk.

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
99% confidence
Finding

The parser accepts --export-env-only and promises not to analyze, but the flag is never checked. When media arguments are supplied, execution still initializes identity and invokes analysis, potentially uploading private footage or incurring service charges despite the user's explicit non-analysis mode.

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 code reads the complete local file and passes its bytes to the analysis service without a visible disclosure or consent gate. This is legitimate analysis functionality, not evidence of malicious exfiltration, but indoor-camera footage may contain sensitive household information beyond the intended plant subject.

Content

No source excerpt is available for this finding.

The documented environment-export-only mode still executes video analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The environment-export-only flag is parsed but never used to alter control flow. With input supplied, the script proceeds to identity initialization and remote analysis, violating the promised no-analysis boundary and potentially transmitting media or consuming paid resources.

Content

No source excerpt is available for this finding.

The documented create operation silently falls back to overwriting existing records.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

Any insertion exception triggers attempts to update an existing record by ID and then username or realname, copying nearly every field, including identity and token fields. A conflicting or malformed input can therefore turn an intended insertion into an unintended overwrite; shared identity storage increases the integrity consequences beyond plant reports.

Content

No source excerpt is available for this finding.

The shared DAO implements personal-profile management unrelated to leaf-aging analysis.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The User model stores real names, birthdays, sex, and age, and UserDao exposes CRUD operations on these records. Although notification delivery may justify a recipient identifier or delivery token, these demographic fields are not justified by the plant-monitoring purpose; DAO initialization also automatically creates and migrates the sys_user table (L185-L215).

Content

No source excerpt is available for this finding.

ai_chat claims to invoke an analysis agent, but execution is disabled and no analysis result is returned.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 explicitly promises a subprocess invocation of the OpenClaw agent, and L33–L34 promise an analysis result or session ID. However, the invocation is commented out, and L68–L70 instead access stderr on an empty dictionary, raising an exception that is caught at L81–L82. Consequently, the method never invokes the documented agent and implicitly returns None rather than performing the promised analysis.

Content

No source excerpt is available for this finding.

FileUtil.open silently truncates existing files.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The generically named open helper opens files in 'w' mode, immediately erasing any existing contents. No docstring, comment, confirmation, or user-facing message discloses this destructive behavior; the exception handler only runs if opening fails.

Content

No source excerpt is available for this finding.

HTTP requests can silently register a remote account and persist its identity and tokens.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

A generic HTTP request can implicitly register a remote account using the current identity and persist returned profile data and tokens locally. This creates undisclosed external identity processing and credential retention that are not inherently necessary for leaf-aging analysis.

Content

No source excerpt is available for this finding.

HTTP requests can silently trigger remote user registration.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The request fallback explicitly sends silent=1 and register=1 to a remote login endpoint, allowing routine HTTP activity to create an account without a distinct confirmation step. This is the same implicit-registration issue identified by SDI-1, not a separate exploit.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

The POST transmits the resolved username as both openId and mobile to the configured remote service during implicit registration. External requests are not inherently unsafe, but here the unconfirmed identity disclosure is concrete, and the username selection can even use API_SECRET_KEY as the identity value.

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 YAML files, including during module import.

Low
Category
Output Handling
Confidence
92% confidence
Finding

YamlUtil.load creates the parent directory and writes a configuration file when the requested path is missing, without a comment, docstring, confirmation, or user-facing log disclosing this side effect. BaseEnum.init_subclass invokes this loader during class definition (L044–L058), so importing the module can modify the filesystem rather than merely read 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