Back to skill

Security audit

Accessibility Tester

Security checks for vulnerabilities and agentic risk

Overview

This looks like an accessibility audit skill, but its runnable code can report perfect WCAG compliance without actually checking the target.

Treat this as a prototype or reference only, not as a compliance gate or audit tool. Before installing for real use, require actual browser/DOM inspection of the supplied target, fail-closed behavior for untested criteria, and output that clearly distinguishes tested, untested, failed, and simulated results.

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

Error
Location
index.ts:286
Finding

Accessibility Audit Returns False Compliance Results Without Inspecting the Target

Content
View full analysis
{ // Simulated audit - in real implementation would use browser automation const results = await this.wcagEngine.testAll({}); const keyboardResults = await this.keyboardTester.test({}); const passed = results.filter((r) => r.status === 'pass').length; const failed = results.filter((r) => r.status === 'fail').length; const score = Math.round((passed / results.length) * 100); return { summary: { score, level: failed === 0 ? 'AAA' : failed <= 2 ? 'AA' : 'A', passed, failed, warnings: 0, }, principles: { perceivable: { score: 100, issues: [] }, operable: { score: 100, issues: [] }, understandable: { score: 100, issues: [] }, robust: { score: 100, issues: [] }, }, issues: { critical: results.filter((r) => r.severity === 'critical'), serious: results.filter((r) => r.severity === 'serious'), moderate: results.filter((r) => r.severity === 'moderate'), minor: results.filter((r) => r.severity === 'minor'), }, compliance: { wcag21A: failed === 0, wcag21AA: failed === 0, wcag21AAA: failed === 0, }, }; } } ``` Supporting fail-open behavior appears in `index.ts:92-99`, where the non-text-content rule starts with an empty collection instead of reading the supplied context: ```typescript async testNonTextContent(context: any): Promise { const elements: TestElement[] = []; const failingEleme ...[truncated 3547 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · SKILL.md (reported line 169)May include surrounding context.

md
const allResults = await this.testAll(context);
    return allResults.filter(result => {
      const rule = this.rules.get(result.ruleId);
      return rule && rule.level === level;
    });
  }

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · SKILL.md (reported line 226)May include surrounding context.

md
// ... Add all 61 WCAG rules
    
    return rules;
  }
  
  private async testNonTextContent(context: TestContext): Promise<WCAGTestResult> {

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The keyboard navigation tester returns hardcoded success values for tab order, focus management, and keyboard interactions without evaluating the page. Because keyboard accessibility is a critical compliance area, these unconditional passes can conceal serious usability barriers for keyboard-only and assistive-technology users.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The audit method claims to assess a provided URL but ignores the target entirely and runs simulated checks against empty objects, producing results that look authoritative. In a security/compliance workflow this can mislead users into believing an application was accessibility-tested when no real analysis occurred, causing missed defects and false compliance attestation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code reports WCAG 2.1 A/AA/AAA conformance based solely on a few simulated checks and sets all conformance flags true whenever no stubbed failures are found. This is dangerous because it can be used as false evidence of compliance in development, procurement, or regulatory contexts despite lacking coverage and real validation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The non-text-content check states it verifies images for missing alt text, but it only filters an always-empty local array and never inspects the provided context. This creates systematic false negatives for a core WCAG criterion, undermining trust in the tool and allowing accessibility defects to pass unnoticed.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

This TypeScript test file exercises tester.audit('http://localhost:3000'), which indicates the skill performs network or HTTP access against a target URL. In this file there is no confirmation prompt, disclosure string, or warning comment explaining that the audit connects to a live endpoint, so the network-touching behavior is not visibly disclosed here.

Content

No source excerpt is available for this finding.