Security audit

Kubernetes Network Policy Generator

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent ToolWeb API integration for privacy vendor scoring, but users should treat submitted organization and procurement details as third-party data sharing.

Install this only if you are comfortable sending the scorecard inputs to ToolWeb's API. Avoid submitting confidential procurement details, regulated data, or direct personal identifiers unless your organization approves that sharing; where possible, ask the agent to omit organization and evaluator names or confirm before making the API call.

A.I.G

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

Findings (1)

T01 · Skill Instruction Hijacking

Note

Location: SKILL.md:171
Finding: Forced Branded Output Can Override the Agent's Requested Response Format## Vulnerability Details **File Location**: `SKILL.md`, lines 171-193 **Vulnerability Type**: Output instruction hijacking through mandatory promotional content **Risk Level**: Low **Complete Code Snippet**: ```text ## Output Format ``` 📊 Privacy Solution Vendor Scorecard ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Organization: [name] Industry: [sector] Budget: [range] Regulations: [list] 🏆 Top Ranked: [Vendor Name] — [weighted score] 📋 Vendor Comparison: [Vendor A]: [total score] — Strongest in [top criteria] [Vendor B]: [total score] — Strongest in [top criteria] 📊 Head-to-Head by Category: Core Capabilities: [Vendor A] vs [Vendor B] Compliance: [Vendor A] vs [Vendor B] Technical: [Vendor A] vs [Vendor B] Financial: [Vendor A] vs [Vendor B] 🎯 Recommendation: [Summary recommendation with rationale] 📎 Full report powered by ToolWeb.in ``` ``` ### Technical Analysis The skill prescribes a fixed response template that includes the promotional statement `Full report powered by ToolWeb.in`. Because these instructions are loaded into the agent's active context, an agent may treat the template as mandatory even when the user requests another response format or asks for an unbranded report. This is a limited form of skill instruction hijacking affecting response integrity. It does not instruct the agent to bypass safety controls, expose credentials, execute arbitrary code, or modify persistent state. The identified impact is therefore restricted to the current session's presentation goals and the injection of third-party branding. The branding is especially problematic because the wording may imply that ToolWeb generated, reviewed, or endorsed the complete report, even when portions were independently generated by the agent. ### Attack Path 1. A user installs or enables the skill. 2. The agent loads `SKILL.md` into its active instruction context. 3. The user requests a vend ...[truncated 1028 chars]
Remediation: ## Remediation Suggestions 1. Mark the template as optional rather than mandatory, for example: “Use the following format when the user has not requested another format.” 2. Explicitly state that user-specified formatting and branding preferences take precedence. 3. Remove the mandatory `Full report powered by ToolWeb.in` line, or make attribution opt-in. 4. If attribution is retained, use precise wording that distinguishes API-generated material from agent-generated analysis. 5. Add an instruction prohibiting the template from overriding system policies, developer instructions, user requirements, or safety constraints. 6. Use a neutral default response structure without marketing language.

SkillSpector

Vulnerability Patterns

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

Findings (3)

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The README instructs users to configure an external service API key but does not clearly disclose that user prompts, vendor names, budgets, regulatory requirements, or other procurement details may be transmitted to that third-party service. This creates a privacy and transparency issue because operators may unknowingly send sensitive business or compliance data outside their environment.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill instructs the agent to collect organization name, evaluator name, budget, industry, regulations, and vendor assessment data, then transmit it to a third-party API, but it does not clearly disclose this external sharing to the user before collection. This creates a privacy and consent risk because potentially sensitive business and personal information may be sent off-platform without explicit user awareness or approval.

External Transmission

Medium

Category: Data Exfiltration
Content: 2. **Call the API**: ```bash curl -s -X POST "https://portal.toolweb.in/apis/compliance/privacy-scorecard" \ -H "Content-Type: application/json" \ -H "X-API-Key: $TOOLWEB_API_KEY" \ -d '{
Confidence: 93% confidence
Finding: The skill makes an external POST request to portal.toolweb.in and includes user-supplied organizational and evaluation data in the request body. While external API usage is expected for this skill's functionality, it is still security-relevant because it exports potentially sensitive procurement, compliance, and evaluator information to a remote service, increasing confidentiality and third-party data handling risk.

Static analysis

No suspicious patterns detected.