atelier-greenfield

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands and package managers like Bun and Maven to initialize project structures, manage dependencies, and run validation gates.- [INDIRECT_PROMPT_INJECTION]: The agent reads local project configuration files (e.g., pom.xml, next.config.ts) to determine the appropriate scaffolding variant, creating a surface where untrusted file content could potentially influence agent instructions.
  • Ingestion points: Local project structure and configuration files in the target directory.
  • Boundary markers: Instructions require user confirmation for the target directory, variant selection, and final commit.
  • Capability inventory: File system writes, shell command execution (Bun/Maven), and Git hook configuration.
  • Sanitization: The skill uses neutral placeholders (e.g. atelier) for identity metadata, but does not explicitly describe input sanitization for external project-provided configuration values.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 04:06 PM
Security Audit — agent-trust-hub — atelier-greenfield