videodb
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes external video transcripts, creating a surface for indirect prompt injection when this data is used in LLM generation prompts. This is inherent to the skill's primary purpose of video perception and memory. Ingestion points include video.get_transcript_text() and the events dumped by scripts/ws_listener.py.
- [DYNAMIC_EXECUTION]: Uses dynamic script generation for SDK interactions through shell heredocs in SKILL.md. This is a standard instructional pattern for the tool and is used locally for legitimate integration.
- [EXTERNAL_DOWNLOADS]: Dependencies on the videodb SDK and python-dotenv are fetched from official registries. As the primary libraries for the skill, these are documented as standard requirements.
- [COMMAND_EXECUTION]: Employs local command execution to manage background listeners and video capture sessions in reference/capture.md and scripts/ws_listener.py. These follow standard daemon management and session recording patterns.
Audit Metadata