seo-image-gen
Warn
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local installation script (
./extensions/banana/install.sh) and several Python scripts (scripts/presets.py,scripts/cost_tracker.py,scripts/generate.py) that are not included in the provided file set, preventing verification of their security impact. It also utilizes shell commands for image processing viamagickandffmpeg. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'banana' extension and an MCP server (
@ycse/nanobanana-mcp) from third-party repositories not associated with the provided author context. - [DATA_EXFILTRATION]: Usage statistics, including raw user prompts and generation metadata, are logged to
~/.banana/costs.json. This local persistence of sensitive user inputs could be accessed by other local processes. - [PROMPT_INJECTION]: The skill processes user-supplied image descriptions and interpolates them into a prompt engineering framework. Ingestion points: User-provided strings in command arguments (e.g.,
<description>). Boundary markers: Missing. Descriptions are directly inserted into the reasoning brief components. Capability inventory: File system writes, shell execution, and network access via tools. Sanitization: None specified for user input.
Audit Metadata