ugc-lifestyle-photos
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include a configuration step to add a Model Context Protocol (MCP) server from
https://mcp.apimage.org/mcp. This involves registering a remote service provided by the vendor to extend the agent's toolset. - [COMMAND_EXECUTION]: The 'Setup' section contains a shell command (
claude mcp add ...) intended to be executed by the user to configure the Claude CLI environment. This command modifies local configuration files to include external transport headers and endpoints. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where untrusted data (external product assets) is ingested and processed by image generation tools.
- Ingestion points:
reference_imagesparameter in thegenerate_imagetool call withinSKILL.md. - Boundary markers: None; the instructions do not specify delimiters or safety warnings to prevent the agent from being influenced by metadata or instructions embedded within processed assets.
- Capability inventory: The skill uses image generation and background replacement tools (
generate_image,create_brand_asset,replace_background) which could be influenced by injected instructions. - Sanitization: There are no explicit instructions for validating or sanitizing the content of
PRODUCT_ASSETbefore it is passed to the generative model.
Audit Metadata