The Agent Skills Directory

[COMMAND_EXECUTION]: Extensive use of npx hyperframes subcommands for project scaffolding, rendering, and cloud infrastructure management via AWS SAM and Terraform.
[EXTERNAL_DOWNLOADS]: The skill downloads a pinned Chrome browser via the browser ensure command and AI coding skills from GitHub during initialization. The capture command also fetches assets from arbitrary user-provided URLs.
[CREDENTIALS_UNSAFE]: Manages authentication by storing API keys and tokens in the vendor-specific ~/.heygen/credentials file.
[INDIRECT_PROMPT_INJECTION]: The capture command ingests content from external websites to scaffold project templates, creating a surface for indirect prompt injection attacks where malicious instructions could enter the agent's context.
Ingestion points: npx hyperframes capture <url> in SKILL.md and references/init-and-scaffold.md.
Boundary markers: Mentions checking for BLOCKED.md to stop processing failed captures, but does not specify prompt delimiters for ingested HTML/JS content.
Capability inventory: File system writes, local/cloud rendering, and deployment to AWS Lambda and Google Cloud Run.
Sanitization: Instructs users not to infer brand/design data from partial files in a blocked capture, but lacks automated sanitization details for captured site data.
[DATA_EXFILTRATION]: The feedback command transmits reproduction packets to a public channel for support. The skill explicitly warns users to anonymize absolute paths, secrets, and identifiers before submission to prevent sensitive data exposure.

hyperframes-cli