The Agent Skills Directory

Command Execution: The skill runs a bundled Python helper script (scripts/pptx_talk_track.py) to parse and update PowerPoint files. This script is designed to run locally using standard libraries, which is a safe approach for file manipulation.
Indirect Prompt Injection: Since the skill ingests content from external files and notes, there is a theoretical surface area for instructions within those files to influence behavior. The skill mitigates this by focusing on text generation and requiring user confirmation for file saves. * Ingestion points: PowerPoint documents, slide images, and text-based outlines provided by the user. * Boundary markers: The instructions emphasize that speaker notes are the primary source of truth, though they do not use specific delimiters for the ingested text. * Capability inventory: The skill can read file contents and write speaker notes to a new PowerPoint file using the provided script. * Sanitization: Content is processed by the model for script generation, with explicit guardrails provided to the agent regarding confidential or sensitive information markings.

presentation-talk-track-builder