ads-photoshoot

Pass

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill utilizes an indirect prompt injection surface by interpolating external data into prompts for the image generation engine.
  • Ingestion points: Product descriptions provided by the user and fields from the local brand-profile.json file are used to build style-specific prompts in Step 4.
  • Boundary markers: The templates (e.g., [product description], professional product photography...) lack delimiters or instructions to treat the interpolated content as untrusted data.
  • Capability inventory: The skill can execute shell commands via the /banana tool and write files to the local filesystem (./product-photos/).
  • Sanitization: No validation or escaping is performed on the ingested metadata before it is used in prompt construction.
  • [COMMAND_EXECUTION]: The skill interacts with the local environment by executing the /banana CLI tool for setup, configuration verification, and image generation. It also accesses specific local file paths for brand identity (brand-profile.json) and cost tracking (~/.banana/costs.json).
Audit Metadata
Risk Level
SAFE
Analyzed
May 5, 2026, 10:42 AM
Security Audit — agent-trust-hub — ads-photoshoot