Back to skill

Security audit

agent-teams

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because it found a medium-severity warning about deleting lock files without first stopping processes that may still be using them; the evidence is incomplete because several referenced paths were missing.

Review the lock-file recovery guidance before installing: deleting a lock file while other processes are active can allow conflicting access to shared state. The scan covered all seven listed components, but its analysis was marked partial because some referenced paths were missing, so the assessment is not complete.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

Manual lock-file deletion is instructed without warning users to stop all processes sharing those locks.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

Unlinking a lock file does not release an fcntl lock held on its inode: an existing process can retain the old lock while another process creates and locks a replacement file, allowing both to enter the critical section. This is particularly dangerous in a concurrently running agent team because it can corrupt shared inboxes or task state and cause duplicate task execution; atomic JSON replacement does not prevent these logical races.

Content

No source excerpt is available for this finding.