The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local Python and Node.js scripts to perform its core functions. It calls scripts from other skills in the environment, specifically for searching prompt inspiration, batch processing images with Gemini models, and taking screenshots of HTML designs.
Evidence: python3 .claude/skills/ai-artist/scripts/search.py, .claude/skills/ai-multimodal/scripts/gemini_batch_process.py, and node .claude/skills/chrome-devtools/scripts/screenshot.js.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to browse Pinterest to research design references and styles.
Evidence: Navigation to pinterest.com to search for and screenshot reference pins for art direction.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it ingests content from external websites (Pinterest) and uses it to guide design generation and script parameters.
Ingestion points: Research results and screenshots from pinterest.com (SKILL.md Step 2).
Boundary markers: No explicit delimiters or instructions are provided to help the agent distinguish between researched design content and potential malicious instructions embedded in those pages.
Capability inventory: The skill can execute local scripts (python, node), write files to the assets/banners/ directory, and automate browser actions.
Sanitization: There is no mention of sanitizing or validating the data retrieved during the research phase before it is used to influence prompts for image generation.

ckm-banner-design