The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article content to identify key concepts for illustration.
Ingestion points: Article content is read via read_file or processed from pasted text in the conversation (SKILL.md, references/workflow.md).
Boundary markers: The instructions lack explicit text delimiters to separate the untrusted article content from the agent's system instructions during analysis.
Capability inventory: The skill generates images via the image_generate tool and uses terminal and write_file for file and network operations.
Sanitization: The skill includes a dedicated 'Pitfalls' section and core principles requiring the agent to 'Strip secrets' (scan for API keys, tokens, or credentials) before writing any output to disk.
[COMMAND_EXECUTION]: The skill uses the terminal tool to perform standard file management and asset downloads.
Evidence: The workflow specifies using curl or wget to download image assets from URLs provided by the image_generate tool (references/workflow.md).
Evidence: Uses common shell commands such as cp and mv for reference management and creating file backups.
[EXTERNAL_DOWNLOADS]: The skill retrieves generated images from external URLs.
Evidence: Downloads images from dynamic URLs generated by the image generation backend using standard network utilities.

baoyu-article-illustrator