Back to skill

Security audit

Plant Night Respiration Rate Analysis | 植物夜间呼吸作用强度估算

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious with high confidence because scanners report account, credential, billing, and general media-analysis capabilities that do not clearly fit its stated plant-respiration purpose.

Consider not installing until the account, credential, and media-upload behavior is clearly explained and you can verify where local media and credentials are sent. The scan is partial, and it reports a missing reference, so the evidence does not cover every referenced item.

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
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
Findings (30)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk implements user-data management rather than thermal-image acquisition, leaf-temperature analysis, CO₂ processing, or respiration estimation. Although database storage can support an imaging application, the concrete schema and specialized operations here concern user identities and authentication data, with no demonstrated connection to the declared plant-monitoring workflow. This constitutes a materially different purpose and undeclared data handling in the supplied code. No network transmission or token use is shown, and the chunk alone does not establish malicious intent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Generic file, JSON, timing, and polling helpers could reasonably support the declared analysis workflow. However, the chunk also implements an undeclared account-registration, credential-management, and billing integration that is materially broader than thermal-image respiration estimation. In particular, the HTTP helper automatically resolves an identity, obtains remote account tokens, and adds credentials to requests without restricting absolute URLs to the configured API host. No camera capture, thermal analysis, CO₂ processing, or respiration inference appears in this chunk. This assessment is limited to the supplied code; the account and billing operations occur when the relevant helpers are invoked, not automatically merely because they are defined.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied code identifies itself as a video-analysis tool and calls skill.get_output_analysis with a video path or URL. It also exposes a history-listing operation through skill.get_output_analysis_list(open_id=...). It contains no visible implementation of thermal-camera acquisition, nighttime monitoring, leaf-versus-air temperature analysis, CO₂ integration, or respiration estimation. The imported skill implementation is not supplied, so its analysis internals cannot be verified; nevertheless, the visible generic video workflow and user-associated history capability materially differ from the declared specialized monitoring purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied chunk does not implement the declared domain-specific behavior. Its agent subprocess execution is disabled, so it should not be characterized as executing external commands. The imported base classes and API service may provide additional functionality, but their implementations are not supplied and cannot establish that the declared workflow exists.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The visible implementation is a general API-backed media-analysis and report-management wrapper rather than the described thermal plant-monitoring workflow. File submission and result polling could support the declared purpose, but nothing in this chunk constrains the analysis to thermal plant imagery or implements the domain-specific processing described. Report-history retrieval is also an additional capability not identified in the description. Imported API implementations are not supplied, so the assessment does not establish what analysis the backend ultimately performs.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The visible code exposes a general API-access interface rather than implementing the declared respiration-analysis behavior. HTTP helpers and pagination could support that workflow, so their presence alone would not establish a mismatch. However, the user-lookup and broad data-management capabilities have no demonstrated relationship to the declared purpose. This assessment is limited to the supplied chunk: it defines callable methods but does not show them being invoked, establish which external resources they access, or prove malicious activity.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding

Generic AI-analysis endpoint configuration could support the declared purpose and is not itself a mismatch. However, the explicit health-order report export endpoint introduces an inconsistent resource not explained by the description. This chunk only configures endpoints; it does not demonstrate actual report access or network requests. The absence of plant-analysis logic in a configuration file alone would not establish a 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-plant-night-respiration-rate-analysis"
description: "In a plant factory, a fixed thermal imaging camera continuously captures thermal images of the plant canopy leaves at night (no-light period), analyzes leaf temperature trends (respiration releases heat, causing leaf temperature to be slightly higher than air temperature), and combined with optional ambient CO₂ sensor data (respiration releases CO₂ raising concentration), an AI model estimates the relative respiration. | 通过植物工厂内的固定热成像摄像头,在夜间(无光照时段)连续采集植物冠层叶片的热图像,分析叶片温度变化趋势�

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

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 local-media analysis workflow does not warn users that their input media is sent to an external API.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The local-file workflow does not clearly state that the file's contents will be uploaded to an external analysis service. References to cloud reports and APIs provide partial context but do not clearly disclose the media transfer, potentially exposing proprietary facility imagery or research data without informed authorization.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

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 advertises --export-env-only as not executing analysis, but the parsed flag is never checked. With an input or URL, execution continues into analysis, potentially transferring media or consuming service credits despite the caller explicitly selecting a 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 submitted to the analysis API without disclosing the upload.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The code reads the complete local media file and passes its bytes to the analysis API without a visible disclosure or authorization gate. File extension and size checks do not protect confidentiality, so apparently local analysis can disclose proprietary plant-factory imagery or research data to an external processor.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to execute analysis, but the code still runs it.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser accepts --export-env-only and promises not to perform analysis, but the program never checks this flag before invoking identity resolution and the analysis or history-listing paths. When an input is supplied, a user expecting a configuration-only operation can unintentionally trigger processing and potentially external transmission of plant-factory footage through the delegated analysis implementation.

Content

No source excerpt is available for this finding.

The service exposes username-based user lookup, an account-related capability unrelated to plant respiration analysis.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

At L098-L101, get_user_by_username forwards a supplied username to RequestUtil.get_user_by_username. The manifest describes thermal-image and optional CO₂ analysis, neither of which justifies retrieving user accounts by username.

Content

No source excerpt is available for this finding.

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

Medium
Category
Output Handling
Confidence
95% confidence
Finding

YamlUtil.load creates parent directories and writes a YAML file when the requested configuration is absent, without a comment, docstring, prompt, or log disclosing this side effect. BaseEnum.init_subclass invokes this loader during class definitions, 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.

The skill includes user-profile and authentication-token database management unrelated to plant respiration analysis.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L448-L473 define a user model containing personal information and authentication tokens and expose it through a DAO with inherited read, update, and delete operations. DAO initialization also creates registered tables and explicitly migrates sys_user (L184-L215), coupling database use to user-account storage. These capabilities are not justified by the manifest's thermal-image and optional CO₂ analysis purpose.

Content

No source excerpt is available for this finding.

The ai_chat documentation claims to invoke an AI agent, but the implementation never executes it or returns an analysis result.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L26 explicitly says the method invokes openclaw agent through subprocess. However, the invocation at L61-L66 is commented out; L68 instead creates an empty dictionary, and accessing result.stderr raises an exception that is caught at L81-L82, causing the method to return None without performing AI analysis.

Content

No source excerpt is available for this finding.

FileUtil.open silently opens files in write mode, truncating existing contents without disclosure.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

Opening a path with mode 'w' immediately truncates an existing file, even before the caller writes anything. This method has no confirmation, user-facing log, docstring, or comment explaining that destructive behavior; exception tracing does not disclose a successful truncation.

Content

No source excerpt is available for this finding.

The request helper silently provisions remote user accounts and persists authentication data, beyond the stated plant-respiration analysis purpose.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

An ordinary API request can automatically register a remote account and persist returned authentication information without a separate enrollment step. More concretely, API_SECRET_KEY takes precedence as the username and can consequently be transmitted as both openId and mobile, unnecessarily propagating a credential into identity fields; the plant-analysis description does not disclose this behavior.

Content

No source excerpt is available for this finding.

HTTP requests can silently trigger remote user registration without disclosing the additional identity transmission.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The helper sets silent and register to 1 and submits an identity to a registration/login endpoint when credentials are missing, making registration a hidden side effect of unrelated API requests. This creates an undisclosed external identity transfer, potentially including an API secret used as the username, and overlaps with the enrollment issue identified in SDI-2.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

An external authentication request is not inherently unsafe, and the supplied file does not establish that the configured endpoint is malicious. However, this call receives API_SECRET_KEY as its identity argument when that setting is present and token acquisition is needed, sending the secret in ordinary account fields where it may be logged or stored as identity data; this is the transmission component of the same enrollment issue.

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"):