The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill is highly susceptible to indirect prompt injection and path traversal attacks because it allows user-provided data to define file paths and content in script executions.
Ingestion points: Filenames (e.g., 'my-deck.md'), slide titles, and author names supplied by the user are used as arguments for the init_presentation.py script in SKILL.md.
Boundary markers: Absent; there are no instructions to the agent to treat these inputs as untrusted or to validate their structure.
Capability inventory: The skill has file modification capabilities and command execution privileges via uv run.
Sanitization: No sanitization or validation logic is mentioned, creating a risk for arbitrary file writes if paths are not restricted.
[COMMAND_EXECUTION] (INFO): The skill uses uv run to invoke internal scripts (init_presentation.py, check_contrast.py) for workflow automation. While standard for skills, the lack of input validation on the arguments passed to these commands elevates the risk level.

slide-creator