historical-science-video-prod
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The script
scripts/generate_video_materials.pytakes user-controlled inputs via CLI arguments (--theme,--era, and--core_conclusion) and interpolates them directly into prompt templates sent to the LLM (e.g., in functionsgenerate_script,generate_character_design, andgenerate_storyboard). The lack of delimiters or instruction-ignore warnings creates a surface for prompt injection where malicious input could redirect the LLM's output content. - [COMMAND_EXECUTION]: The skill requires the execution of a local Python script (
scripts/generate_video_materials.py) to function. This script performs network operations to reach an LLM provider and writes multiple files to the local file system based on the LLM's generated content. - [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user data and lacks proper defensive measures.
- Ingestion points: The
theme,era, andcore_conclusionarguments inscripts/generate_video_materials.pyare the primary entry points for untrusted data. - Boundary markers: Absent. The prompt templates in
scripts/generate_video_materials.pydo not use clear delimiters to separate user data from system instructions. - Capability inventory: The script has the ability to create directories (
os.makedirs) and write files (open().write()) to the local disk based on generated content. - Sanitization: Absent. There is no evidence of input validation or output sanitization before the data is used in prompts or written to files.
Audit Metadata