The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the core CLI package from the NPM registry via npx and downloads a version-pinned Chromium binary for deterministic rendering. It also retrieves AI coding skill updates from the vendor's GitHub repository.\n- [COMMAND_EXECUTION]: Utilizes shell commands for project scaffolding (init), environment health checks (doctor), and infrastructure deployment to cloud providers using tools like AWS SAM and Terraform.\n- [DATA_EXFILTRATION]: Uploads project archives to the HeyGen cloud API for managed rendering and sends anonymous telemetry, which includes environment fingerprinting to identify agent and sandbox runtimes.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content through multiple ingestion points, including web content capture from arbitrary URLs, external JSON variable interpolation, and automated media transcription. While it includes boundary markers like schema validation and non-interactive flags, the ingested data interacts with capabilities such as shell command execution and network operations, presenting a vulnerability surface for indirect prompt injection.\n- [REMOTE_CODE_EXECUTION]: Downloads and executes code via npx and performs remote rendering by zipping and uploading local project code to HeyGen infrastructure.

hyperframes-cli