The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves ingesting and processing external, untrusted data (articles and blog posts) to create visual metaphors and generate image prompts. This represents an indirect prompt injection attack surface where a maliciously crafted article could contain instructions designed to override the agent's behavior or hijack the image generation process.
Ingestion points: SKILL.md (Step 1: Digest the Article) instructions the agent to read and identify anchor points from user-provided text.
Boundary markers: The instructions lack explicit delimiters (e.g., XML tags or clear separators) or specific "ignore embedded instructions" warnings for the ingested article content.
Capability inventory: The agent is instructed to use generate_image tools or emit prompts for image generation based on the processed content.
Sanitization: There are no instructions for sanitizing, escaping, or validating the input article content before it is used to populate the prompt template.

article-illustrations