The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation and evaluation cases include examples of using standard command-line tools such as cwebp, mogrify, and jpegoptim for image compression, as well as curl for diagnostic audits of image assets on a website. These are standard utilities for the described marketing tasks.
[CREDENTIALS_UNSAFE]: The skill instructions in SKILL.md prompt the agent to ask the user for API keys for third-party image generation services (e.g., Gemini, Flux, Ideogram). This is a functional requirement for using these tools and does not involve hardcoded credentials or automated data exfiltration.
[PROMPT_INJECTION]: The skill ingests data from project-level context files such as .agents/product-marketing.md. While this constitutes an indirect prompt injection surface, it is a standard practice for project-aware agents and lacks high-privilege execution capabilities that would make it a significant risk.
Ingestion points: .agents/product-marketing.md (identified in SKILL.md).
Boundary markers: No explicit markers are used to delimit content from the ingested context files.
Capability inventory: Execution of image processing shell commands and diagnostic network requests.
Sanitization: No specific sanitization or validation logic is implemented for the ingested marketing context.

image