video-script

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from external URLs and local repositories during its research phase, establishing an indirect prompt injection surface. This is an inherent risk for research-based agents but is mitigated as the skill's output is restricted to text generation for human review.\n
  • Ingestion points: External articles and code repositories processed via WebFetch and Read/Grep/Glob (SKILL.md).\n
  • Boundary markers: Not present.\n
  • Capability inventory: Local file writing and directory creation (SKILL.md).\n
  • Sanitization: None specified.\n- [DATA_EXFILTRATION]: The skill prompts users for promotional links and social media handles, which it stores in a local memory file (~/.claude/projects/.../memory/video-promo.md). This allows the information to persist across sessions for automated insertion into video metadata. This functionality is transparent and limited to user-provided public information.\n- [COMMAND_EXECUTION]: The skill performs standard file system operations to create date-based project directories and save markdown files. These actions are confined to the user's workspace and are necessary for organizing video production assets.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 01:32 PM