The Agent Skills Directory

[DYNAMIC_EXECUTION]: The shell script synthesize-audio.sh uses the source command to load provider-specific logic from the scripts/tts-providers/ directory based on the PRESENTATION_TTS environment variable or command-line arguments. This pattern of dynamic loading from computed paths is an execution surface. Additionally, the Node.js script extract-narrations.ts utilizes dynamic import() to load generated narrations.ts files from chapter directories during the build process.
[COMMAND_EXECUTION]: The skill automates project setup and validation by executing several external commands. The scaffold.sh script runs npm ci and npm run check. The check.mjs script executes npx tsc to perform TypeScript validation. The TTS synthesis runner invokes curl, jq, and potentially external CLI tools such as ffmpeg or mmx (MiniMax CLI) to process media.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided content from plan.md (in YAML format) and .srt subtitle files to generate code, documentation, and animation timing data. This ingestion of untrusted external data serves as a potential surface for influencing agent behavior. While the skill includes a defensive check.mjs script to enforce design and code constraints, the structural ingestion of external text should be managed with caution.
[EXTERNAL_DOWNLOADS]: The skill fetches assets and interacts with well-known services. The fonts.css file imports Google Fonts, and various TTS provider scripts (openai.sh, azure.sh, gcloud.sh) perform network requests to official APIs. These operations are essential for the skill's primary functions of media rendering and audio synthesis.

web-video-presentation