creativeclaw-create-avatar
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data, including character briefs and personal photos, to generate prompts for external image and voice models. While the instructions emphasize preserving user-approved details and avoiding unauthorized inventions, the ingestion of external text and media creates a standard theoretical surface for indirect prompt injection. The skill mitigates this by requiring manual inspection of results and explicit confirmation steps.
- Ingestion points: User-supplied photos and character briefs (SKILL.md, references/avatars/identity.md).
- Boundary markers: Instructions explicitly tell the agent to "Preserve approved likeness," "Preserve exact quoted copy," and "Never infer cloning permission."
- Capability inventory: The skill uses tools for image generation (
generate_image), speech synthesis (generate_speech), and media manipulation (merge_media). - Sanitization: The agent is instructed to "Inspect text, identity, hands, product geometry... before calling the result finished."
- [EXTERNAL_DOWNLOADS]: The documentation references several well-known technology and AI services, including Google (ai.google.dev), ElevenLabs (elevenlabs.io), Cartesia (cartesia.ai), xAI, and HeyGen. These are used for legitimate API documentation and service integration in line with the skill's purpose. All referenced domains are established and recognized technology providers.
- [COMMAND_EXECUTION]: The skill interacts with an MCP server hosted at
https://app.creativeclaw.co/mcp/chatgpt, which belongs to the skill's author. The tools provided are restricted to media management and generation tasks. No arbitrary shell command execution or unauthorized file system operations were detected.
Audit Metadata