The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied articles and reference images to generate prompts and analysis files. There is a risk that malicious instructions embedded in these articles could influence the agent's behavior during the illustration process.
Ingestion points: SKILL.md (Step 2: Analysis of article content), references/workflow.md (Vision analysis of reference images).
Boundary markers: None explicitly defined in the prompt templates to distinguish between article data and agent instructions.
Capability inventory: terminal (curl, cp), write_file, image_generate, vision_analyze.
Sanitization: Includes a 'strip secrets' step to remove credentials from source content, but does not specifically sanitize against prompt injection patterns.
[COMMAND_EXECUTION]: The skill instructs the agent to use terminal commands for file management and network operations.
Evidence: references/workflow.md instructs the use of cp to copy local reference images and curl -sSL -o to download generated images from URLs.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to download generated images from remote URLs provided by image generation tools.
Evidence: references/workflow.md Step 6: 'Download the returned URL to {output-dir}/NN-{type}-{slug}.png via terminal (e.g. curl -sSL -o ... {url})'.

baoyu-article-illustrator