creativeclaw-gpt-image-2
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill integrates an external Model Context Protocol (MCP) tool to handle media and asset tasks.
- Evidence:
agents/openai.yamlreferences thecreative-clawtool hosted athttps://app.creativeclaw.co/mcp/chatgptwhich belongs to the skill vendor. - [INDIRECT_PROMPT_INJECTION]: The skill instructions describe a workflow that processes untrusted external data in the form of image URLs and user-supplied prompts.
- Ingestion points: The skill utilizes
image_url,extras.image_urls, and specific prompt segments defined inSKILL.md(Outcome, Scene, References, Visible text, etc.). - Boundary markers: There are no specific delimiters or instruction-override protection markers defined for the interpolation of user-provided data into the model prompts.
- Capability inventory: The skill's capabilities are focused on media and image generation via the
creative-clawtool referenced inagents/openai.yaml. - Sanitization: No explicit content filtering, URL validation, or prompt sanitization procedures are provided within the skill's instructional framework.
Audit Metadata