The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses a sensitive local authentication file at ~/.codex/auth.json. This file is used to manage credentials for the image generation provider. While documented as part of the vendor tool's functionality, accessing local credential files is a security-sensitive operation.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface.\n
Ingestion points: The skill processes external data including user-provided topics, audience, and tone, as well as imported reference templates (PDF, PPTX) and content fetched from web videos via yt-dlp. (File: SKILL.md)\n
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when the agent processes this external content.\n
Capability inventory: The skill can execute various system commands via slides-grab and yt-dlp, write files to the local system, and make network requests to AI providers (OpenAI, Google) for image generation. (File: SKILL.md)\n
Sanitization: No evidence of sanitization or validation of the input data is provided before it is interpolated into prompts for image generation or slide layout.\n- [EXTERNAL_DOWNLOADS]: The skill includes external resources from well-known services: fetches the Chart.js library from https://cdn.jsdelivr.net/npm/chart.js and uses yt-dlp to download video content from the web into the local assets directory.\n- [COMMAND_EXECUTION]: The skill relies on executing a suite of CLI tools (slides-grab subcommands) and yt-dlp to perform its core functions of generating, validating, and exporting slide decks.

slides-grab-html