skills/vllnt/skills/build-prototype/Gen Agent Trust Hub

build-prototype

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to query the npm registry and install numerous packages from the @vllnt scope, including @vllnt/ui, @vllnt/logger, and @vllnt/analytics. The instructions in references/vstack/capabilities/project-alignment/references/default-packages.md require the agent to fetch the newest stable versions at runtime.
  • [COMMAND_EXECUTION]: The workflow requires executing package installation commands (e.g., via npm or yarn) and running the generated prototype artifacts to observe behavior and error paths.
  • [DYNAMIC_EXECUTION]: The skill creates a 'runnable artifact' for a vertical slice of a web, server, or CLI application. This dynamically generated code is intended to be executed locally to validate product hypotheses.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data to define the 'journey, hypothesis, constraints, and iteration budget' in SKILL.md. This data is then used to generate executable code, creating an injection surface.
  • Ingestion points: User-provided request for resolving problem, journey, and hypothesis in SKILL.md.
  • Boundary markers: None explicitly defined in the instructions to separate user input from the agent's code generation logic.
  • Capability inventory: File system writes (generating prototype code), npm package installation, and execution of the resulting artifact.
  • Sanitization: The skill relies on a 'Quality validation' protocol in references/vstack/protocols/quality-validation.md which includes review and critique, but does not specify automated sanitization or escaping of the input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 04:54 PM