The Agent Skills Directory

[SAFE]: The skill utilizes Markdown backticks strictly for formatting local file paths (e.g., references/style-dna.md) and referencing the built-in image_gen tool. These are static documentation markers and do not trigger dynamic context injection or shell command execution.
[SAFE]: All visual assets, style guidelines, and character definitions are stored as static Markdown or local image files. No external network requests, unverified package installations, or remote script executions are present.
[SAFE]: The skill does not access sensitive environment variables, system configuration files, or user credentials. Its file operations are limited to organizing generated image assets in a local directory.
[INDIRECT_PROMPT_INJECTION]: The workflow involves processing user-provided articles to create illustration prompts. While the skill lacks explicit delimiters for this external data, the downstream capability is limited to image generation, which does not present a mechanism for privilege escalation or environment compromise. The inherent risk is low.
[SAFE]: Metadata and instructional content were reviewed for obfuscation, homoglyphs, and hidden instructions. All content, including Chinese-language prose, is readable and consistent with the skill's stated purpose.

ian-xiaohei-illustrations