Back to skill

Security audit

Rose Pest & Disease Detection | 月季/玫瑰常见病虫害识别

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because scanners report account and credential handling, user-history features, and unexpected remote activity beyond its stated rose-diagnosis purpose.

Consider avoiding installation until the account, profile, and token handling is clearly explained, the export-only behavior and media-upload disclosures are addressed, and the development HTTP configuration is reviewed. The Skillspector scan was partial because it reported a missing local reference, so some evidence is incomplete.

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
skills/smyx_common/scripts/config-dev.yaml
Finding

Development endpoints expose authentication tokens over HTTP

Content
View full analysis

Impact

An on-path attacker can capture and replay tokens within their validity and granted permissions, and observe or modify unencrypted analysis traffic.

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 (26)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Generic database access could be a supporting implementation detail, but this chunk specifically implements user-profile and token persistence rather than storage of images, diagnoses, severity grades, or treatment suggestions. It contains no image analysis or pest-detection logic, and the description does not disclose user-account management. The supplied chunk therefore has a materially different purpose. This assessment is limited to the shown code; it does not establish that the broader skill lacks rose-detection functionality or that any stored data is transmitted externally.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code explicitly identifies itself as a video-analysis tool and exposes video input and history-listing operations. It contains no visible rose-specific detection logic, pest categories, severity grading, or control recommendations. The imported skill implementation is not supplied, so its analysis domain cannot be confirmed; however, the visible video-oriented interface and identity-linked history retrieval materially differ from the declared image-based rose diagnostic capability. Optional output-file writing is a supporting implementation detail, not a mismatch by itself.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

Ordinary formatting, polling, filesystem, and HTTP helpers can support image analysis and are not mismatches by themselves. However, this chunk also implements an undisclosed account and billing integration: it obtains a persistent identity, can register that identity remotely, persists returned credentials, and injects authentication and identifying metadata into outgoing requests. These behaviors exceed the declared rose-detection purpose. The supplied chunk contains no image-analysis or diagnosis implementation, although its role as a shared utility module means that absence alone does not establish a mismatch. Most effects occur when these helpers are called; the chunk does not establish that they run during every diagnosis.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The visible behavior does not implement the declared diagnostic purpose, and its only substantive method is a nonfunctional generic chat helper. No subprocess is actually launched, so the commented command should not be treated as an executed capability. Imported base classes and services are not supplied; they could contain additional functionality, but this chunk alone does not substantiate the description.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The media-analysis path is broadly consistent with the description, but the user-linked history retrieval mode and pet-specific controls extend beyond it. Identity resolution alone could be supporting infrastructure; the separate history-access capability is the substantive mismatch. The actual detection implementation is imported and not supplied, so this chunk cannot establish whether rose diseases, severity grades, or control suggestions are actually produced.

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-rose-pest-disease-detection-analysis"
description: "AI-powered pest & disease detection for roses (Rosa spp.). From garden cameras or mobile phone images of leaves, young shoots and flower buds, detects common rose enemies including black spot (black round/irregular spots with yellow halo), powdery mildew (white powdery layer on leaves/shoots), spider mites (tiny red/white dots on leaf back with webbing in severe cases) and aphids (green/black clustered tiny insects on shoots and buds). Outputs pest/disease type, severity grade and general control suggestions. Helps gardeners detect issues early and act in time. Scenarios: home ga

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 that local images or videos are sent to an external API.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The workflow directs automatic processing of local media without clearly warning that its contents leave the device for external analysis. Garden-camera footage and phone photos can contain people, property details, or location metadata, so a horticultural purpose does not eliminate privacy risk.

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 parser accepts --export-env-only and promises no analysis, but the program never checks the flag before identity initialization or analysis. When supplied with media, an invocation intended to be non-executing can therefore upload data and potentially incur service charges.

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 analysis API without disclosure of the upload.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The code reads the entire local media file and supplies its bytes to the analysis API without a visible consent or disclosure gate. Rose photographs and garden videos can include private surroundings or embedded metadata, making unexpected external transfer a meaningful privacy exposure.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to run analysis, but the code ignores it and still invokes 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 history flag supplied, the command can perform remote operations despite its explicit promise not to run analysis, exposing media or causing unexpected account activity.

Content

No source excerpt is available for this finding.

The service exposes username-based user lookup, a capability unrelated to rose pest and disease analysis.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

The get_user_by_username method delegates a supplied username to a user-lookup operation. Looking up user accounts is not a direct requirement of analyzing rose images or producing severity grades and control suggestions, and the manifest does not declare account-management functionality.

Content

No source excerpt is available for this finding.

The skill includes persistent user-profile and credential management unrelated to rose pest and disease detection.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The user model includes bearer-token fields as ordinary SQLite strings, alongside personal profile information, without visible secret-storage protections. Persisting such records in a shared workspace database can expose reusable credentials to other processes or skills with filesystem access; the rose-analysis description does not disclose this additional sensitive-data footprint.

Content

No source excerpt is available for this finding.

FileUtil.open silently truncates existing files without disclosing its write-only behavior.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The generically named open helper always uses mode 'w', which immediately erases an existing file's contents. No docstring, explanatory comment, confirmation, or user-facing log discloses this destructive behavior; exception tracing does not warn about successful truncation.

Content

No source excerpt is available for this finding.

The request helper silently provisions remote user accounts and persists user records, beyond the declared rose-image analysis functionality.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

An ordinary HTTP request can trigger remote login or registration using silent=1 and register=1, followed by local persistence of returned user information and tokens. This creates undisclosed account and credential-management side effects beyond the declared image-analysis task, without an explicit consent gate in the supplied helper.

Content

No source excerpt is available for this finding.

The HTTP helper silently transmits an identity for automatic remote login or registration.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The helper automatically sends the selected identity as both openId and mobile to a remote login/registration endpoint. That identity can originate from workspace configuration or an API-key file, so a routine rose-analysis request can disclose an identifier or sensitive configured value without a separate user decision.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

This POST concretely transmits the selected user identity and source metadata to the configured account service, with automatic registration enabled. External requests are not inherently unsafe, but this request performs undisclosed identity transfer and account creation rather than simply submitting rose images; the configured destination's trustworthiness is not established by this source.

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"):

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

Low
Category
Output Handling
Confidence
92% confidence
Finding

YamlUtil.load creates missing parent directories and writes a YAML file without a warning, log, comment, or docstring explaining this side effect. BaseEnum.init_subclass invokes this loader during class definition (L044–L059), so importing the module can modify the filesystem even when the caller only expects to 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

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:

ai_chat claims to invoke the OpenClaw agent, but its agent invocation is disabled.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 explicitly says this method invokes the OpenClaw agent through subprocess, but the entire invocation is commented out at L61–L66. Instead, it creates an empty dictionary and accesses result.stderr, raising an exception that is caught and suppressed; the method therefore performs no agent analysis and returns None.

Content

No source excerpt is available for this finding.