Back to skill

Security audit

Flowering & Fruit Set Rate Analysis | 番茄/辣椒开花坐果率分析

Security checks for vulnerabilities and agentic risk

Overview

This plant-analysis skill is not clearly malicious, but it needs Review because it sends user media to cloud APIs while silently creating or reusing a persistent identity and token store.

Review before installing. Use this skill only if you are comfortable with uploaded media and report history being handled by the listed cloud service, and with the skill creating or reusing a local identity plus cached tokens for that service. Avoid sensitive images or private URLs unless the publisher provides clearer privacy, retention, and account-control terms.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (65)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

Automatic remote authentication, user creation/login, local workspace access, and network communication are highly sensitive behaviors that are not central to a simple image-analysis skill and are not prominently disclosed. In context, this is especially dangerous because home users may upload private local images believing they are only being analyzed, not linked to accounts and sent through a broader service stack.

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-flowering-fruit-set-rate-analysis-analysis"
description: "AI-powered flowering and fruit-set rate analysis for tomato / chili plants. From home grow-box or mobile phone images of flowering/fruit clusters, uses object-detection models to count open flowers (fully-opened corolla with visible stamens) and successfully-set young fruits (enlarged ovary, ~0.5-1cm green baby fruits), and computes fruit-set rate = young fruits / flowers × 100%. Helps growers evaluate pollination, nutrition and environmental adaptability, and guides hand-assisted pollination or water/fertilizer adjustment. Scenarios: home smart grow-boxes, greenhouses, balcony

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill accepts remote video URLs and local video files even though its declared purpose is image-based flowering/fruit-set analysis. This scope mismatch is dangerous because it broadens the attack surface unexpectedly: users may supply arbitrary remote resources or oversized/complex media to a backend that was not clearly intended or disclosed for such processing, increasing privacy, abuse, and backend handling risks.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The script materially contradicts the skill’s stated purpose: it presents itself as flowering/fruit-set image analysis but actually implements generic video analysis and history retrieval. This kind of capability mismatch is dangerous because users may submit unintended data types or trust the skill with data handling and processing flows they did not consent to, creating a deceptive-skill/supply-chain risk.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The model stores token and open_token values even though authentication handling is not justified by the plant image analysis use case. Persisting authentication-like secrets in a local SQLite database creates a high-value target for credential theft, replay, or cross-component abuse if the file is read by other skills, users, or processes.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The code resolves an open_id by reading local files, reusing database identities, or silently generating and persisting a new default user for later remote authentication. That behavior is not justified by the advertised image-counting function and can bind the skill to a persistent remote identity without the operator's informed consent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

This shared utility performs remote identity provisioning, token retrieval, and token persistence that are unrelated to tomato/chili flowering or fruit-set image analysis. In a skill whose stated purpose is local/agronomic image analysis, hidden account bootstrapping and credential handling materially expands the trust boundary and can cause undisclosed remote linkage of user/workspace identity.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill declares no explicit tool/permission scope even though its documentation describes shell execution, local file handling, network access, and implicit identity use. This weakens containment and makes it harder for a host platform to enforce least privilege, increasing the chance of unintended data access or remote calls.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The documentation expands the skill from image analysis into cloud-backed historical report retrieval and report-link generation, which introduces additional data-access and privacy implications. This scope growth is not inherently malicious, but it is security-relevant because historical records are often more sensitive than a single analysis result.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill does not clearly warn users that local files or supplied URLs may be sent to a remote cloud API for analysis. This is a significant privacy and consent issue, especially for home users uploading local media from personal devices.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The default trigger fires on general uploads of tomato, chili, eggplant, or similar flower/fruit images or videos without clear confirmation of user intent. Over-broad triggers can lead to unintended local file processing or remote upload of media the user did not mean to analyze with this cloud-backed skill.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The history-query trigger phrases are broad and lack exclusion conditions, which can cause automatic retrieval of cloud-linked historical reports from casual conversation. In a skill that uses implicit identity association, that increases the chance of exposing prior analysis data unintentionally.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Silent use of internal identity parameters and fallback/default-user creation is not justified by the stated gardening purpose and can cause cross-user data exposure if the wrong identity is reused. Automatic identity creation and reuse also undermine informed consent and auditability for cloud-stored reports.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file documents pet health analysis endpoints even though the skill is presented as a plant flowering and fruit-set analysis tool. This kind of cross-domain mismatch can cause the agent or integrators to call unrelated backend APIs, potentially exposing, misrouting, or mishandling data and creating unsafe operational behavior through configuration confusion.

Content

No source excerpt is available for this finding.