The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts and system tools to manage the media workflow.
The compress.py script invokes ffmpeg and ffprobe via subprocess.run. These calls use list-based arguments, which is a secure implementation that avoids shell injection vulnerabilities.
The agent is instructed to use curl and jq to interact with the OpenRouter API for model discovery and cost tracking.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading media files and API metadata from openrouter.ai.
All network operations target a well-known service and are authenticated via user-provided keys. The skill downloads generated media content and polls job statuses over HTTPS.
[SAFE]: The skill exhibits several security best practices throughout its design.
Credential Handling: Instructions explicitly forbid hardcoding API keys and direct the user to secure environment variable storage, specifically mentioning .env files.
Output Sanitization: The gallery generation script (build_gallery.py) correctly uses html.escape on all API-provided data (such as prompts and model IDs) before rendering it into the local HTML gallery, preventing potential Cross-Site Scripting (XSS).
Cost Protection: The skill enforces budget agreement and sequential generation to prevent runaway API costs, protecting the user's financial resources.

generate-media