The Agent Skills Directory

[COMMAND_EXECUTION]: The skill dynamically discovers and executes local scripts for image generation based on environment variables or directory traversal. In references/codex-imagegen.md, it searches for packages/baoyu-codex-imagegen/src/main.ts and executes it via bun, and also honors the $BAOYU_CODEX_IMAGEGEN_BIN environment variable for execution.
[EXTERNAL_DOWNLOADS]: The skill uses npx -y bun as a fallback mechanism to execute the image generation wrapper, which involves downloading the bun package from the public NPM registry.
[PROMPT_INJECTION]: The skill incorporates untrusted article data into image generation prompts, representing an indirect prompt injection risk. Ingestion points: User-provided article content processed during the analysis phase (detailed in references/workflow.md Step 2). Boundary markers: None identified in the prompt templates; labels are interpolated without explicit instructions to the generator to ignore embedded commands. Capability inventory: Shell command execution for backends and image generation via the Skill tool. Sanitization: No sanitization or escaping of the interpolated article text is described in Step 5.1 of references/workflow.md.

baoyu-article-illustrator