character-design
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-provided character descriptions and reference images, which are then interpolated into prompts for external AI generation tools.
- Ingestion points: The SKILL.md file defines a workflow for collecting character types, identity anchors, styles, and reference images from users.
- Boundary markers: The skill utilizes a structured CHARACTER ANCHOR and SHOT VARIABLE template system in references/anchor-system.md and references/prompt-patterns.md to organize inputs, which helps distinguish identity traits from scene instructions.
- Capability inventory: The skill calls the genmedia CLI tool to perform network operations with various AI providers (including OpenAI, Fal.ai, and ByteDance) and handles file system actions, including uploading reference images and downloading generated outputs to ./outputs/characters/.
- Sanitization: While the instructions emphasize staying within the character anchor and using schema-supported fields, there are no specific guidelines or functions for sanitizing or escaping the user-provided text and image metadata before it is passed to the generation models.
Audit Metadata