hyperframes-cli

Warn

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The npx hyperframes skills command fetches and installs agent skills directly from GitHub to the local authoring environment.
  • [EXTERNAL_DOWNLOADS]: The skill downloads external assets including a bundled Chromium browser (npx hyperframes browser ensure) and a machine learning model for on-device search.
  • [DATA_EXFILTRATION]: The npx hyperframes feedback utility sends detailed reproduction packets, including command history and error logs, to a public channel. Users must manually anonymize paths and credentials.
  • [DATA_EXFILTRATION]: The skill implements environment fingerprinting via telemetry, detecting specific AI agent runtimes and sandbox environments by checking for environment variables and system files.
  • [INDIRECT_PROMPT_INJECTION]: The npx hyperframes capture command creates projects from external URLs (Ingestion point: references/init-and-scaffold.md). The skill lacks boundary markers or sanitization for captured content, while maintaining a high capability inventory including command execution via render and check.
  • [CREDENTIALS_UNSAFE]: The skill interacts with sensitive file paths including ~/.heygen/credentials and ~/.aws/credentials, and uses environment variables like FIGMA_TOKEN.
  • [COMMAND_EXECUTION]: The cloudrun deploy and lambda deploy subcommands utilize Terraform and AWS SAM to provision cloud infrastructure, involving the creation of IAM roles and security policies.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 9, 2026, 06:49 AM