Back to skill

Security audit

solve-tech-problems

Security checks for vulnerabilities and agentic risk

Overview

This is a simple response-formatting skill for engineering problem solving, with a scope wording mismatch but no executable code, persistence, credential access, or hidden behavior.

Installers should treat this as a materials-engineering advice template and avoid relying on it for unrelated technical domains unless the wording is corrected. The publisher should narrow the title and body text to materials engineering to match the declared scope.

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
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest explicitly scopes the skill to materials engineering, but the body instructs the agent to solve general technical problems. This creates a scope-expansion vulnerability: routing or policy layers may trust the manifest while the prompt behavior encourages use outside the approved domain, increasing the chance of unsafe guidance in unrelated high-risk technical areas.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The inline instructions contradict the declared domain-only restriction by directing the model to identify and implement solutions to broad technical problems. Such contradictions weaken safety controls because the model may follow the more permissive operational instructions, enabling out-of-scope assistance that could include sensitive engineering, industrial, or dual-use topics.

Content

No source excerpt is available for this finding.