The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The browser ensure command downloads a pinned version of the Chrome browser to ensure rendering reproducibility. The capture command downloads images, SVGs, and other assets from user-supplied URLs. Additionally, asset preprocessing tools like transcribe and remove-background may download machine learning models (e.g., Whisper) on first run.
[REMOTE_CODE_EXECUTION]: The lambda deploy command builds and deploys code to AWS Lambda. The validate and inspect commands execute the project's HTML and JavaScript within a headless Chrome instance to perform runtime checks and layout audits.
[COMMAND_EXECUTION]: The skill makes extensive use of system commands, including npx, ffmpeg, ffprobe, aws sam, and bun to manage the video production pipeline.
[DATA_EXFILTRATION]: The telemetry feature collects anonymous usage data, which includes fingerprinting the runtime environment (e.g., detecting if it is running in Docker, gVisor, or Firecracker) and identifying the specific AI agent environment (e.g., Claude Code, GitHub Copilot, Devin) by checking for vendor-specific environment variables.
[CREDENTIALS_UNSAFE]: The AWS Lambda integration (hyperframes lambda) requires access to sensitive cloud credentials, typically stored in ~/.aws/credentials or provided via environment variables, to provision and manage infrastructure.
[INDIRECT_PROMPT_INJECTION]: The capture command ingests content from external websites, creating an attack surface where malicious instructions embedded in a target web page could theoretically influence the agent's behavior during the project conversion process.

hyperframes-cli