Back to skill

Security audit

Claude Devfleet

Security checks for vulnerabilities and agentic risk

Overview

This skill openly describes a multi-agent coding workflow with automatic dispatch and merges, and its sensitive behavior is purpose-aligned rather than hidden.

Install this only if you intend to let a local DevFleet service coordinate multiple coding agents. Review the generated mission plan before approving dispatch, and use protected branches or manual review settings if you do not want completed agent work merged automatically.

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
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill describes a workflow where mission completion triggers automatic merge and dependent missions are automatically dispatched, but it does not prominently warn the user that unreviewed agent-generated code may be merged and further agents may continue operating on that state. In a multi-agent coding system with full tool access and isolated worktrees, this can propagate erroneous or unsafe changes quickly and reduce opportunities for human review before code is integrated.

Content

No source excerpt is available for this finding.