Back to skill

Security audit

assess-automotive-lighting-ip

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent for patent/design screening, but it can send potentially confidential product images to an external PatSnap service without requiring explicit user consent or confidentiality checks.

Review before installing if you handle unreleased designs, CAD exports, customer materials, NDA-covered images, or export-controlled content. Use it only when third-party PatSnap processing is authorized, and require explicit confirmation of which images will be sent, whether multiple views are needed, and whether redaction or metadata removal is required.

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
SKILL.md:211
Finding

External transmission of potentially confidential product images lacks a mandatory consent and data-handling gate

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 211–235
Vulnerability Type: Sensitive information disclosure to a third-party service
Risk Level: Medium

Vulnerable Code Snippet

text
Use the official Patsnap Patent Research MCP server for live FTO execution.

Official page: https://open.patsnap.com/marketplace/mcp-servers/patsnap-ip-searching

Configuration key: `patsnap_patent_research`.

Transport: `streamableHttp`.

Current Connect-panel pattern verified on 2026-08-07:

`https://open.patsnap.com/marketplace/mcp-servers/patsnap-ip-searching`

Always copy the current URL from the official Connect panel.

Never place a real API key in the report, skill, source control, or logs.

Use `fto_review` for invention-patent FTO screening.

Use country, application-date, legal-status, and assignee filters as appropriate.

Use `search.mode: lite` for a clearly labeled preliminary screen.

Use `search.mode: pro` when the task and available access justify deeper analysis.

Use `design_fto` for design-risk screening from one product or design image.

Remember that only the first `input.images` item is used by that tool.

Select the most representative view for the tool call.

Run separate calls for materially different views when necessary.

Technical Analysis

The Skill instructs the Agent to use an external PatSnap MCP service for live freedom-to-operate analysis and to submit a representative product or design image to design_fto. Automotive drawings and product images may reveal confidential geometry, unreleased styling, dimensions, program details, or trade secrets.

This network operation is relevant to the declared patent and design-right screening functionality, and the destination is represented as PatSnap's official service. The instructions also prohibit placing API keys in reports, source control, or logs. Therefore, the reviewed content does not demonstrate credential theft or intentional malicious exfiltration.

Howeve ...[truncated 2766 chars]

Remediation
View remediation

Remediation Suggestions

  1. Add a mandatory pre-transfer confirmation that identifies:

    • The specific external service.
    • The files or images to be transferred.
    • The purpose of the transfer.
    • Whether multiple views will be submitted.
  2. Require explicit user consent before the first external submission. Consent to perform an analysis should not automatically be treated as consent to disclose confidential artifacts to a third party.

  3. Add a confidentiality gate requiring the Agent to determine whether submitted material is:

    • Public.
    • Internal.
    • Confidential.
    • Subject to an NDA.
    • Export-controlled or otherwise restricted.
  4. Require confirmation that third-party processing is authorized under applicable organizational policy and contractual obligations.

  5. Enforce data minimization:

    • Crop images to the feature needed for comparison.
    • Redact labels, dimensions, program names, and unrelated geometry.
    • Strip EXIF and other metadata.
    • Avoid transmitting additional views unless necessary and separately approved.
    • Prefer reduced-resolution copies when sufficient.
  6. Require disclosure or verification of relevant service policies, including retention, deletion, model-training use, data residency, access controls, and subprocessors.

  7. Provide a safe fallback for restricted material, such as a documented search strategy, locally prepared feature description, or user-executed search plan that does not upload the original image.

  8. Record only a minimal audit entry indicating authorization and the submitted artifact category. Do not log image contents, credentials, or raw confidential service responses.

  9. Preserve the existing prohibition on embedding API keys in reports, source control, and logs.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 275)May include surrounding context.

md
Use `intelligent_image` for available patent drawings.

Do not rely on translated text without checking the original when wording is outcome-determinative.

### Optional: Advanced Patent Search

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 673)May include surrounding context.

md
Examples may include halo-like light rings, hammer-shaped daytime-running lights, star motifs, continuous light bars, pixel patterns, or other recognizable signatures.

Do not attribute a motif to a brand without verification.

Do not claim that the motif is patented or registered without an identifier and source.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instruction Use lang="en". mandates a specific language/locale for the output report. Under the policy, forcing a language without user opt-in is a natural-language policy violation unless the constraint is explicitly justified as region-specific, which this file does not provide.

Content

No source excerpt is available for this finding.