ra-video-title

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a clear, localized workflow for generating video titles. It relies on internal logic and local references to suggest titles and update a specific markdown file (视频标题.md) using standard file system tools.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data and maintains file-writing capabilities.
  • Ingestion points: The skill reads video scripts, project context files (e.g., narration_segments.json), and user-supplied benchmark titles.
  • Boundary markers: None are explicitly implemented to distinguish between processed data and agent instructions.
  • Capability inventory: The agent is instructed to use apply_patch to write the final title to 视频标题.md.
  • Sanitization: No specific sanitization or filtering is performed on the ingested text.
  • Analysis: While this creates a potential surface for instructions embedded in scripts to influence the agent, the risk is mitigated by the skill's strict formatting requirements (8-12 two-part candidates) and the narrow scope of its file-writing capability.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 08:46 AM
Security Audit — agent-trust-hub — ra-video-title