The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted academic papers in PDF format as its primary data source, creating an attack surface for indirect prompt injection. Ingestion points: Extracted paper content in paper-content.md is used as the basis for the Narrative IR and Slide IR generation steps. Boundary markers: Absent. The instructions do not direct the agent to use delimiters or 'ignore embedded instructions' warnings when including the extracted paper content in its generation prompts. Capability inventory: The skill executes shell commands via scripts/build.sh (latexmk, xelatex) and manages environments with uv. It also performs various file system operations to create the presentation structure. Sanitization: No explicit sanitization or filtering of the extracted text is performed before it is interpolated into the prompts.
[COMMAND_EXECUTION]: The skill relies on shell command execution via subprocess.run and a dedicated bash script (scripts/build.sh) to invoke the LaTeX build system and the uv package manager. These operations are fundamental to the skill's purpose of generating and compiling Beamer presentations from source text.
[EXTERNAL_DOWNLOADS]: The skill utilizes the docling library and the uv package manager, which download document processing models and standard software dependencies from well-known external registries and services.

paper2beamer