short-drama-video-prompts
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external files including storyboards, screenplays, and character descriptors to generate final video prompts. This creates a potential surface where malicious instructions embedded in the input data could be interpolated into the final prompt. The risk is mitigated by the skill's instruction to use markdown blockquotes ('>'), explicit section headers (### 可复制提示词), and XML-style delimiters () for verbatim content, which helps differentiate instructions from data. \n
- Ingestion points: The skill reads from JSONL files such as
shots.jsonl,keyframes.jsonl, andscreenplay-index.jsonl. \n - Boundary markers: Employs markdown blockquotes and tags like
<d>for dialogue. \n - Capability inventory: The skill is capable of generating prompt files that are intended for subsequent AI video generation tasks. \n
- Sanitization: Relies on structural formatting and explicit labeling rather than active filtering. \n- [SAFE]: The Python scripts (
container_check.py,motion_timing_check.py,music_spec_check.py) are focused on local data validation and arithmetic checks for timing. They use standard libraries and do not perform network operations or access sensitive system files. \n- [SAFE]: External references to repositories (e.g., MiniMax-AI GitHub) are documented neutrally and target well-known technology service providers, posing no risk to the skill's environment.
Audit Metadata