Skills
skills/study8677/antigravity-workspace-template/agent-repo-init/Gen Agent Trust Hub

agent-repo-init

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script executes git init using subprocess.run with a static argument list. This is a standard operation for repository initialization and is implemented safely to prevent shell injection.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes external inputs (CLI arguments) that influence filesystem operations.
  • Ingestion points: scripts/init_project.py accepts --project-name, --destination-root, and --template-root as CLI arguments.
  • Boundary markers: No specific delimiters are used for these inputs within the script's logic.
  • Capability inventory: The script uses shutil.copytree to copy directories, Path.write_text to create files, and subprocess.run to execute git commands.
  • Sanitization: The project_name is validated against a strict alphanumeric regex. Path inputs are resolved using .resolve(), and a check is performed to ensure the destination is not within the template root.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 05:11 AM