Back to skill

Security audit

Subclip Text and Captions

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because a scanner found that cloud transcription may send audio off-device without a user-facing warning or confirmation; the review covered all listed components.

Review the skill before installing if your media may contain sensitive audio. Consider whether you are comfortable with cloud processing when speaker detection is requested, since the scanner reports that the skill does not require a warning or confirmation first. The verdict is suspicious, not malicious; the scan reports complete coverage of the two listed components, but that does not guarantee the skill is 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
  • 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)

Cloud transcription is authorized without a user-facing warning that media audio will leave the device.

Medium
Category
Output Handling
Confidence
87% confidence
Finding

The skill permits cloud transcription when speaker detection is requested, without requiring disclosure that audio will leave the device or confirmation that the user accepts remote processing. Requesting speaker labels is not necessarily informed authorization to upload potentially sensitive recordings. Explicit cloud selection and support for no-cloud processing reduce the risk, but do not close this consent gap for speaker-aware transcription.

Content

No source excerpt is available for this finding.