Back to skill

Security audit

Attack Surface Mapper

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed security-reporting helper that reads local security logs and writes local coverage reports, with some reporting-accuracy bugs but no evidence of hidden, destructive, or exfiltrating behavior.

Install only if you are comfortable letting the skill read local `.security/` red-team, firewall, anomaly, and audit artifacts and write reports under `.security/surface-map`. Treat its output as advisory until the inter-agent mapping and detection-status bugs are fixed, because it may underreport or misclassify some coverage gaps.

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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
index.js:99
Finding

Inter-Agent Attack Surface Is Silently Omitted from Coverage Analysis

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
index.js:61
Finding

Boolean Detection Results Produce Contradictory Coverage Status

Content
View full analysis
d.action?.message?.includes('Prompt injection')); } // Simplified: Check anomaly logs for 'data-theft' if (vector === 'data-theft') { return blueTeamDetections.some(d => d.metric === 'data_volume' && d.classification === 'CRITICAL'); } // Simplified: Check if config-hardener audit report exists for 'config-security' if (vector === 'config-security') { return auditReports.some(r => r.includes('config-hardener')); } return null; // Default to not detected }, /** * Determine coverage status */ getCoverageStatus(redTested, blueDetected) { if (redTested && blueDetected) return 'COVERED'; if (redTested && !blueDetected) return 'GAP'; if (!redTested && blueDetected) return 'PARTIAL'; // Blue detected, but no formal red test yet return 'GAP'; // No red test, no blue detection }, ``` ### Technical Analysis `Array.prototype.some()` returns a primitive boolean. When a detection exists, `blueDetected` is therefore `true`. The coverage-status function treats that value as a successful detection and may return `COVERED`. The presentation logic separately evaluates `blueDetected.full`. A primitive boolean has no meaningful `full` property, so this exp ...[truncated 1725 chars]
Remediation
View remediation
Vulnerability Patterns
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

YARA rule 'exploit_framework': Exploit framework components and payloads [hacktools]

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 · index.js (reported line 113)May include surrounding context.

js
},

  /**
   * List all known attack surfaces
   */
  listAttackSurfaces() {
    return [
      'CHANNELS',
      'SKILLS',
      'TOOLS',
      'MODELS',
      'MEMORY',
      'INTER-AGENT',
      'SUPPLY_CHAIN'
    ];
  },

  /**
   * Get attack vectors for a given surface
   */
  getAttackVectorsForSurface(surface) {
    switch (surface) {
      case 'CHANNELS': return ['prompt-injection', 'social-engineering', 'phishing'];
      case 'SKILLS': return ['malicious-instructions', 'data-theft', 'privilege-escalation'];
      case 'TOOLS': return ['command-injection', 'path-traversal', 'ssrf', 'rce'];
      case 'MODELS': return ['prompt-injection', 'model-confusion', 'jailbreak'];
      case 'MEMORY': return ['memory-poisoning', 'persistence', 'false-context'];
      case 'INTER_AGENT': return ['agent-to-agent-attack', 'lateral-movement'];
      case 'SUPPLY_CHAIN': return ['typosquatting', 'compromised-package', 'repo-takeover'];
      default: return [];
    }
  },

  /**
   * Loa

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill lists the surface as 'INTER-AGENT' but handles it in getAttackVectorsForSurface() as 'INTER_AGENT', so that attack surface receives no vectors and is silently excluded from coverage analysis. This creates a blind spot in a security mapping tool, causing underreporting of inter-agent attack paths and potentially misleading defenders into believing coverage is complete when it is not.

Content

No source excerpt is available for this finding.