The Agent Skills Directory

[COMMAND_EXECUTION]: Uses system commands such as lsof to detect active development servers and the Vercel CLI for deploying decks.
[EXTERNAL_DOWNLOADS]: Fetches required software including Playwright, python-pptx, and ImageMagick from official registries to support snapshot and conversion functionality.
[DYNAMIC_EXECUTION]: Generates and executes temporary Python scripts for extracting PowerPoint content and Node.js scripts for browser-based snapshots of rendered decks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from PowerPoint files without sanitization or boundary markers, which could allow embedded instructions to influence the agent's code generation.
Ingestion points: Reads external .pptx files in ppt mode via the script in references/ppt-conversion.md.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore commands within the extracted slide text.
Capability inventory: Includes file system writes (page.tsx), command execution (deployment via Vercel), and dependency installation.
Sanitization: Absent; extracted text is mapped directly into React primitives and code templates without filtering.
[DATA_EXFILTRATION]: Provides an automated path to upload generated content to Vercel for public hosting, which involves sending local deck data to a remote service.

slide-deck