Back to skill

Security audit

agent-yurt

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill benign with medium confidence: the scans report no direct AIG findings, while the single YARA match was judged expected for the purpose, though the review is incomplete.

The verdict is benign, but confidence is medium and the scan is incomplete: several local references could not be resolved. Consider whether those references matter to your use, and review the skill’s instructions before installing; benign does not mean risk-free.

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
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • 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)

YARA rule 'info_stealer': Information stealer patterns (credential harvesting, browser data theft) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 51)May include surrounding context.

md
suspicions. Choose the reported severity (`info`, `low`, `medium`, `high`, `critical`) from the observed impact. The service derives the account and reporting agent from your credential.

Include safe response request/operation IDs, component/tool name, error code, time and concise expected/actual behavior when available. Exclude raw logs, private memory, conversation, source files, credentials, cookies and customer payloads. Redact before submission; automatic pattern rejection cannot establish that arbitrary text is safe. Use only the documented context fields. Reports stay private for operator triage.

Report one occurrence after bounded troubleshooting for ordinary transient errors; expected quota waits or a known missing scope are not incidents. Avoid repeated submissions for the same unresolved observation. Retain the exact arguments and stable `idempotencyKey` for an identical retry, respect the shared daily/retained submission caps, and stop retrying if access or the reporting