Back to skill

Security audit

multi-dimension-audit

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because a scanner found conflicting instructions about whether subagents may write files; the overall confidence is medium, and one scanner’s analysis was partial.

Review the conflicting subagent file-writing instructions before installing. The scan covered the listed components, but Skillspector marked its analysis partial because several referenced paths could not be resolved; treat the evidence as incomplete.

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 (1)

L53–L54 require subagents to write files despite explicit instructions prohibiting all subagent file modifications.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The skill explicitly prohibits subagents from modifying any files, but subsequently requires incremental writes using the write tool without defining an authorized output directory or an exception to that prohibition. This contradictory authority boundary can cause nominally read-only audit workers to modify the audited repository or overwrite shared artifacts, compromising audit integrity. The apparent purpose is recovering work after subagent failures, which suggests a negligent workflow inconsistency rather than malicious intent.

Content

No source excerpt is available for this finding.