interactive-theatre-designer
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes systems that ingest untrusted audience input to drive narrative and technical outcomes, creating an indirect prompt injection surface.
- Ingestion points: Audience voting via a web interface (
references/live-systems-tech.md) and data from RFID/BLE tracking hardware (SKILL.md). - Boundary markers: The example implementation lacks delimiters or instructions to distinguish untrusted audience data from system commands.
- Capability inventory: The architecture includes a
ShowControllerthat executes environmental changes, technical cues, and performer instructions based on input. - Sanitization: No evidence of input validation or sanitization is present in the provided design patterns.
- [EXTERNAL_DOWNLOADS]: The technical reference documentation identifies several third-party dependencies required for the live performance systems.
- Evidence: Code snippets in
references/live-systems-tech.mdreferencepython-osc,flask,flask-socketio, andRPi.GPIO.
Audit Metadata