better-t-stack
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the create-better-t-stack CLI via npx and manage dependencies using bun, npm, or pnpm across all setup examples.
- [REMOTE_CODE_EXECUTION]: The instructions involve executing remote code by running the latest version of the create-better-t-stack package directly from the NPM registry.
- [EXTERNAL_DOWNLOADS]: The skill facilitates downloading the CLI tool and cloning project source code from GitHub, both of which are well-known and standard services for developer tools.
- [DATA_EXFILTRATION]: The CLI tool performs telemetry collection by default, sending project configuration and environment metadata to a remote R2 bucket. This is documented as a feature, and instructions are provided to disable it via environment variables.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it directs the agent to interpolate user-provided strings, such as project directory names, directly into shell commands. Evidence: 1. Ingestion points: User-provided project names and stack options in SKILL.md. 2. Boundary markers: Absent; user inputs are mixed with CLI flags. 3. Capability inventory: Shell command execution via npx and git across all setup examples. 4. Sanitization: No instructions are provided for the agent to validate or sanitize user input before command construction.
Audit Metadata