The Agent Skills Directory

[COMMAND_EXECUTION]: Uses standard shell utilities like find and mkdir to manage project directories and locate configuration files such as design-profile.json within the local workspace.
[EXTERNAL_DOWNLOADS]: Executes the agent-slides tool via uvx, which downloads the package from the Python Package Index (PyPI) at runtime. This package is the primary engine used for all extraction, build, and audit tasks described in the pipeline.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from user-provided briefs and PowerPoint templates.
Ingestion points: User-provided content descriptions, topics, outlines, and .pptx template files (specifically in the EXTRACT and BUILD steps).
Boundary markers: None; the instructions do not specify the use of delimiters or instructions to ignore embedded commands within the external content.
Capability inventory: The skill has the ability to execute shell commands via uvx, perform file system operations (read/write), and generate executable operations (fixes_ops.json) to modify documents.
Sanitization: Employs several validation gates including preflight, lint, and qa to verify that the output meets technical contracts and visual standards, though these are not specifically designed to detect malicious instructions.

slides-full