scaffold-cli

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes shell commands to set up the project environment, including initializing Git repositories (git init), installing dependencies (pnpm install), and building the project (pnpm run build).\n- [EXTERNAL_DOWNLOADS]: Downloads and executes the ultracite toolchain manager from the npm registry using pnpm dlx ultracite@latest init as part of the automated setup process.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined strings (like project descriptions) into generated code. \n
  • Ingestion points: Project variables (name, description, repo) gathered in Step 1.\n
  • Boundary markers: Absent in generated source and config files.\n
  • Capability inventory: File system write operations in Steps 2-5; shell execution (git, pnpm, npm) in Steps 6-8.\n
  • Sanitization: Absent during template interpolation.\n
  • Mitigation: Includes security-focused patterns in references/agent-friendly-cli.md, such as containedPath to prevent path traversal and assertSafeId for input validation in the generated software.\n- [DATA_EXFILTRATION]: Promotes secure secret management by configuring GitHub Actions to use OIDC trusted publishing for npm releases, eliminating the need for hardcoded or long-lived API tokens.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 11:43 PM