The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The capture command allows the agent to ingest content from arbitrary URLs, converting them into local project files. For more info:
Ingestion points: The npx hyperframes capture <URL> command (found in SKILL.md and references/init-and-scaffold.md) ingests external web content.
Boundary markers: The skill instructs the agent to look for a BLOCKED.md file as a hard stop for safety violations but does not specify delimiters for the captured data itself.
Capability inventory: The skill has access to shell execution, file system writes, and network operations including cloud infrastructure deployment (lambda deploy, cloudrun deploy).
Sanitization: The hyperframes check utility is used to audit the resulting project for runtime errors and layout defects before rendering.
[EXTERNAL_DOWNLOADS]: The skill downloads several components required for its operation from the vendor's infrastructure.
Fetches a pinned version of Chromium for consistent cross-platform rendering results (npx hyperframes browser ensure).
Downloads project templates and examples from the official registry during the init process.
Installs updated instruction sets ("skills") for AI tools from the vendor's GitHub repository.
[COMMAND_EXECUTION]: Extensive use of CLI tools is required for the full development and deployment loop.
Executes the hyperframes CLI via npx for core functionality.
Orchestrates system utilities such as ffmpeg, ffprobe, docker, and bun.
Invokes cloud management tools including terraform, gcloud, and aws-sam for infrastructure tasks.
[DYNAMIC_EXECUTION]: Deployment commands involve the creation and execution of dynamic code artifacts.
The lambda deploy command uses bun to build a ZIP handler for AWS Lambda.
The cloudrun deploy command builds container images for deployment to Google Cloud Run.

hyperframes-cli