The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes an indirect prompt injection surface as it processes untrusted novel text from a user-controlled source directory. This text is passed through multiple agents for transformation into scripts and prompts.
Ingestion points: Novel text files uploaded to the projects/{name}/source/ directory.
Boundary markers: None explicitly defined in the top-level workflow for separating user content from instructions.
Capability inventory: File system access (creating and writing to project subdirectories), project metadata management (JSON), calling external slash-command skills for image and video generation, and executing Python code via a local manager library.
Sanitization: Not explicitly described in the workflow logic.
[COMMAND_EXECUTION]: The skill provides Python code snippets and instructions for using the lib.project_manager library to manage project states and metadata. This is a vendor-provided resource (ArcReel) used for the skill's primary functionality.
[COMMAND_EXECUTION]: The workflow involves the automated use of ffmpeg for video splicing and composition. This is a standard utility execution for the stated purpose of video generation.

manga-workflow