The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it uses untrusted user content to generate complex visual descriptions and prompts for further AI processing.
Ingestion points: User-provided text or Markdown source files (saved as source.md).
Boundary markers: Employs Markdown code blocks in the base prompt template to wrap source content, providing limited isolation.
Capability inventory: Possesses capabilities to execute shell commands (npx), perform local file writes (analysis.md, storyboard.md, .png, .pdf), and interface with sibling image generation skills.
Sanitization: No explicit sanitization or instruction-filtering of the user input is documented before it is interpolated into prompts.
[COMMAND_EXECUTION]: The skill executes local commands as part of its intended operation.
Uses npx -y bun to invoke a sibling image generation skill (tuzi-image-gen) located at a relative path.
Recommends the use of system-level image optimization tools such as pngquant, optipng, or sips for file size reduction.
Runs a local TypeScript utility (scripts/merge-to-pdf.ts) via the Bun runtime to compile generated comic pages into a final PDF document.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx (Node Package Runner) for inter-skill communication and utility execution, which may trigger the download of packages from the npm registry (a well-known service) if dependencies are missing at runtime.

tuzi-comic