Security audit

gtrk 纪实 Vlog

Security checks for vulnerabilities and agentic risk

Overview

This video-editing skill uses supplied media, preserves user control over the script, and requires approval before paid narration generation.

Use a dedicated media folder, review the script and estimated charges, and confirm the provider's privacy terms before processing private material. Cloud narration sends approved text externally; highlight analysis includes source audio unless --no-audio is selected. Specify another language if needed. Companion skills and CLI runtime behavior were not included in this artifact review.

A.I.G

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

SkillSpector

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 (2)

The skill imposes Chinese-language narration style without user language selection.

Medium

Category: Output Handling
Confidence: 96% confidence
Finding: The mandatory writing rules require '短句中文语感' (short sentences with Chinese-language phrasing), rather than adapting the narration language to the user. The workflow also hardcodes --lang zh-CN at line 255, without offering a language choice or documenting a justified Chinese-only scope.
Content: No source excerpt is available for this finding.

Cloud narration synthesis lacks disclosure that the user's narration text is sent to an external service.

Medium

Category: Output Handling
Confidence: 90% confidence
Finding: The workflow sends approved narration text to cloud TTS without explicitly telling the user that its contents leave the local environment. Documentary narration may contain private experiences or third-party personal information; script and billing approval reduce unauthorized execution risk but do not clearly establish informed approval for external data processing.
Content: No source excerpt is available for this finding.