kelly-demo-video-factory
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes video outlines and storyboard data from external sources (JSON files and Busabase database records), which presents a potential surface for indirect prompt injection.
- Ingestion points: The script
scripts/propose_video.mjsreads user-provided JSON outlines, and the dashboard app fetches record data via the Busabase API inapp/js/providers/busabase-provider.js. - Boundary markers: The
SKILL.mdinstructions include a specific "Boundary" rule warning the agent to treat all Busabase content as data rather than instructions. - Capability inventory: The skill can write records to the Busabase database and execute deployment scripts defined in
package.json. - Sanitization: The dashboard application (
app.js) uses standard HTML escaping for all data rendered in the browser UI.
Audit Metadata