skills/kevinzai/commander/ccc-design/Gen Agent Trust Hub

ccc-design

Warn

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [CREDENTIALS_UNSAFE]: The screenshots skill instructions direct the agent to prompt the user for application usernames and passwords. This sensitive data is subsequently used in a generated Playwright script to authenticate with target applications for capturing private pages.
  • [DATA_EXFILTRATION]: The design-consultation skill implements a lifecycle management framework ('gstack') that collects telemetry. This framework captures skill usage, session details, and project metadata including repository names and branch information, transmitting it to a remote logging system.
  • [REMOTE_CODE_EXECUTION]: The screenshots skill dynamically creates and executes a temporary Node.js script (screenshot-script.mjs) to perform automated browser interactions. The design-consultation preamble also uses a source <(command) pattern to dynamically execute command output.
  • [COMMAND_EXECUTION]: Multiple skills perform local command execution. The web-artifacts-builder executes initialization and bundling scripts. The design-consultation skill runs local binaries for update checks, session management, and configuration management from specific local paths.
  • [EXTERNAL_DOWNLOADS]: The skill suite fetches external dependencies and runtimes. The design-consultation skill initiates the installation of the Bun runtime via its official script and manages the installation of various frontend packages from the NPM registry via pnpm and npm.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 21, 2026, 12:55 AM