stage-design
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows a standard operational procedure for classroom generation without using external network calls, sensitive file access, or obfuscated code.\n- [INDIRECT_PROMPT_INJECTION]: The skill instructions define an attack surface by ingesting user input and external materials for lesson generation. This is inherent to its primary purpose and does not constitute a malicious finding.\n
- Ingestion points: User input from
ask_userand content frommaterialFacts(SKILL.md).\n - Boundary markers: None explicitly defined in the instructions to prevent the model from obeying instructions within the materials.\n
- Capability inventory: The skill has capabilities to create stages, generate scenes, and read/modify stage content (
create_stage,generate_scene,patch_stage,read_stage).\n - Sanitization: No sanitization steps are defined for the input materials.
Audit Metadata