The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage environment setup and media processing.
Runs npm install and npx playwright install chromium for environment setup.
Executes python3 to run local scripts lib/narrate.py and lib/add-accent.py for audio synthesis and dictionary management.
Uses ffmpeg for audio/video muxing.
[EXTERNAL_DOWNLOADS]: The skill initiates several external downloads for dependencies and assets.
Installs Python packages: torch, omegaconf, numpy, ruaccent, transformers, num2words, and runorm.
Downloads Chromium browser binaries via Playwright.
Downloads AI models for the Silero TTS engine into a local cache.
References and provides commands to download a large Russian word database from a third-party repository (danakt/russian-words).
Connects to an unofficial Microsoft Edge API for high-quality voice synthesis.
[REMOTE_CODE_EXECUTION]: The skill generates and executes executable scripts locally.
Translates Gherkin .feature files into executable Playwright .spec.ts files and runs them via npm test or npm run video.
Executes local Python scripts (lib/narrate.py) that perform file operations and audio processing.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface by processing external data.
Ingestion points: Processes Gherkin (.feature) files to define test logic.
Boundary markers: None explicitly mentioned to separate untrusted Gherkin content from the generation process.
Capability inventory: Writing and executing .spec.ts files, executing shell commands for video/audio processing, and performing network operations for TTS.
Sanitization: No evidence of sanitization or validation of the Gherkin steps before they are used to generate Playwright scripts.

video-walkthrough