The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell scripts (scripts/init-project.sh) to automate project setup and executes standard commands like npm install, npx tsx, and npx remotion render. These actions are necessary for building and exporting the video content based on user input.
[EXTERNAL_DOWNLOADS]: Fetches media assets from Tenor and connects to reputable AI platforms like ElevenLabs and HuggingFace (Qwen TTS) to generate high-quality voiceovers. These integrations are standard in modern AI-assisted media workflows and use well-known service providers.
[CREDENTIALS_SAFE]: The skill adopts industry-standard security practices by using .env files for managing API keys (ElevenLabs) and explicitly warning users against committing or sharing these sensitive configuration files.
[INDIRECT_PROMPT_INJECTION]: As the skill summarizes and processes external notes or web content to generate video scripts, it possesses an inherent attack surface for indirect prompt injection. This is a common characteristic of LLM-based content generation tools.
Ingestion points: Processes text, URLs, and SRT files provided by the user in the initial analysis phase.
Boundary markers: No explicit markers or system-level delimiters identified to segregate user content from processing instructions.
Capability inventory: Executes shell commands, performs network requests to AI APIs, and writes generated project files to the local file system.
Sanitization: Input content is interpreted by the LLM (specified as Sonnet in instructions) to generate structured video data; no explicit programmatic sanitization is performed on raw input strings before code generation.

remotion-shorts