linkfox-echotik-batch-video-detail
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill introduces a surface for indirect prompt injection by fetching and rendering untrusted external content.
- Ingestion Points: Untrusted data enters the agent context through the
videosarray returned by the/echotik/batchVideoDetailAPI endpoint processed inscripts/echotik_batch_video_detail.py. Specifically, user-generated fields like the video description (videoDesc) are brought directly into the prompt context. - Boundary Markers: There are no clear boundary markers or isolation wrappers around the fetched text fields to instruct the model to disregard any embedded instructions or behavioral overrides.
- Capability Inventory: The skill includes network connectivity to specified vendor endpoints (
tool-gateway.linkfox.com,api.linkfox.com,agent-api.linkfox.com), local file writing for caching and data storage under thelinkfox/project directory, and presentation workflows. It does not possess advanced execution privileges or arbitrary command-spawning capabilities. - Sanitization: Apart from length truncation for display layout preservation, no scanning, escaping, or strict validation is carried out on text values before they are integrated into the output template.
Audit Metadata