Back to skill

Security audit

doc-reviewer

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a document review skill with one concern: an instruction may allow document metadata to be changed despite a report-only restriction, so ClawScan classifies it as suspicious.

Review the conflicting metadata instruction before installing. Consider whether you are comfortable with a reviewer potentially changing `last-verified` without separate authorization; the overall verdict is suspicious, not a claim of malicious intent.

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)

L27 instructs the reviewer to modify document metadata despite L08's explicit no-edit restriction.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The instruction to update last-verified conflicts with the skill's explicit report-only restriction unless edits are authorized. A tool-enabled reviewer could interpret this specific instruction as an exception and modify document metadata without permission, potentially affecting downstream freshness or governance checks. The scope is limited to verification metadata, and the skill otherwise emphasizes reporting rather than editing, which reduces the risk.

Content

No source excerpt is available for this finding.