The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user-provided content to generate comic storyboards and prompts, which could be exploited for indirect prompt injection attacks.\n
Ingestion points: External text is read from user-specified file paths or direct text input (saved to source.md).\n
Boundary markers: While the skill uses markdown templates, it does not employ explicit delimiters or system instructions to ignore potential commands within the user data.\n
Capability inventory: The skill is capable of writing files to the disk, executing local TypeScript scripts, and calling other agent tools for image generation.\n
Sanitization: No validation or escaping is applied to the input source content prior to processing by the language model.\n- [COMMAND_EXECUTION]: The skill executes local scripts and utility commands to manage configuration and process comic assets.\n
Evidence: It runs scripts/merge-to-pdf.ts via the bun or npx runtime to generate the final PDF output.\n
Evidence: It invokes shell commands such as test -f for file detection and sips or pngquant for image compression and conversion.

baoyu-comic