video-prompting-guide
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates that interpolate external data into prompts for AI video models, creating a surface for indirect injection within SKILL.md.\n
- Ingestion points: User-supplied values for prompt formula components (Shot Type, Subject, Action, Setting, Lighting, Style, Technical) processed in SKILL.md.\n
- Boundary markers: Absent. The prompt examples in SKILL.md do not utilize delimiters or specific instructions to ignore embedded commands within user data.\n
- Capability inventory: The skill uses the Bash(belt) tool to execute AI model inference via the inference.sh platform as defined in SKILL.md.\n
- Sanitization: No evidence of data sanitization or validation is present in SKILL.md to handle potentially malicious input strings.\n- [EXTERNAL_DOWNLOADS]: The skill references platform-specific CLI tools and related skill packages for installation.\n
- Evidence: Mentions npx skills add belt-sh/cli and references installation instructions hosted on raw.githubusercontent.com within SKILL.md.\n- [COMMAND_EXECUTION]: The skill provides instructions for using a specialized CLI tool to interact with AI services.\n
- Evidence: SKILL.md includes examples for belt login and belt app run to perform authenticated model execution.
Audit Metadata