Back to skill

Security audit

Plant Root Health Analysis (Transparent Pot) | 植物根系健康状况(透明盆)

Security checks for vulnerabilities and agentic risk

Overview

The skill can analyze plant-root media, but it also silently creates or reuses an account identity, stores tokens locally, and uses cloud history/report APIs, so it needs review before installation.

Install only if you are comfortable sending plant/root images or URLs to the LifeEmergence cloud service and having the skill silently create or reuse an account-linked identity. Review the provider, retention expectations, and local `data` directory token storage before using history/report features, especially on shared workspaces.

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 (62)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill is presented as a narrowly scoped plant-root vision analyzer, but the described behavior includes broader generic backend access, history/report retrieval, identity linkage, and non-domain-specific parameters. This mismatch is dangerous because users and orchestration systems may grant trust and permissions based on the benign description while the skill can perform materially different data access and remote operations.

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

The metadata triggered a poisoning-related rule, suggesting the manifest may contain suspicious formatting or content patterns that can interfere with tooling interpretation or hide behavior in descriptors. In context with the other mismatches and undeclared capabilities, suspicious metadata should be treated as a trust-reduction signal because manifest text is what reviewers and orchestrators rely on for safe scoping.

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: "smyx-root-health-transparent-pot-analysis"
description: "AI-powered plant root health analysis from transparent pots or smart seedling boxes. Uses fixed cameras to capture images/videos of plant roots, identifies root tip color (white=active, brown=aging, black=rotten), root hair density, branching structure, and detects root rot symptoms (softness, mucus, blackish-brown color). Outputs a root health score (0-100) and vitality grade (Healthy/Normal/Weak/Rotten). Helps early detection of root issues (overwatering rot, fertilizer burn, pathogen infection) and guides care adjustments. Scenarios: smart seedling boxes, transparent pots, pla

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The User model stores identity and credential-adjacent fields such as email, token, and open_token despite the manifest describing only plant root health analysis. Collecting and persisting such sensitive data without functional justification materially raises privacy and credential exposure risk, especially because the same generic DAO supports broad CRUD operations over that data.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The code reads identity values from workspace files and creates persistent default local identities when none are provided, despite the skill's declared purpose being root image analysis. This hidden persistence can bind executions to a reusable account, enabling silent tracking, unintended attribution, and future authenticated requests under a user identity the operator did not knowingly choose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The utility layer performs remote account creation/login and token bootstrapping through /sys/phoneLogin, which is unrelated to plant root analysis and silently establishes external identities. In the context of a camera-based plant-health skill, hidden identity provisioning and authentication management materially increase risk because the skill can transmit user-linked data and obtain service access without clear user consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The HTTP helper automatically attaches user identifiers and authentication tokens to outbound requests without a visible user-facing disclosure or feature-specific justification. For a plant root analysis skill, covert transmission of identity-bearing headers and request fields is especially concerning because the declared functionality does not imply account federation or background authenticated API use.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill declares powerful capabilities in its workflow—shell execution, local file save/read, environment use, and network/API access—but does not explicitly scope or constrain those tools in the manifest. That makes the effective privilege boundary unclear and increases the chance that an agent executes unintended local or network actions when processing ordinary user requests.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill claims to perform visual-only root-health analysis, but also normalizes cloud history lookup and report-link generation as part of operation. This expands the data surface from single-input analysis to account-linked historical data access, which can expose prior records unexpectedly if users or callers believe the skill only inspects the current image/video.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The role definition explicitly says '仅输出基于视觉的评估' and not to provide specific treatment advice. Elsewhere, the skill declares '养护方向建议', structured reports, and report-link output, which goes beyond a visual-only assessment and creates an internal documentation contradiction about intended outputs.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

An overly broad trigger phrase for history-report retrieval can cause the skill to invoke cloud record lookup during ordinary conversation that was not intended as a data-access request. This increases the chance of unprompted retrieval of account-linked historical reports, which is a privacy and least-surprise problem.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill states that uploaded attachments are automatically saved as local files, but does not provide a clear user-facing warning at the point of use. Silent persistence of uploads can retain sensitive media longer than expected and expands exposure to local compromise, accidental reuse, or cross-task leakage.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Automatic identity association and fallback local-user creation introduce account binding and persistent identity state unrelated to simple image analysis. If invoked in shared or misconfigured environments, this can cross-link data to the wrong user, create silent accounts, and expose or contaminate another user's analysis history.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill accepts network URLs and notes that the API service will fetch them remotely, but this behavior is not surfaced as a clear user warning. Remote retrieval can leak access patterns and may be abused to fetch attacker-controlled or internal targets if the backend lacks strong URL validation, creating privacy and potential SSRF-style risk.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The API documentation is clearly inconsistent with the stated purpose of the skill: it describes pet health analysis endpoints and scenario codes rather than plant root health analysis. This is dangerous because it can cause the skill to call the wrong backend services, expose unrelated data domains, or lead to accidental cross-domain data access and privacy/security mistakes during integration.

Content

No source excerpt is available for this finding.