backend-setup

Pass

Audited by Gen Agent Trust Hub on Oct 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads project templates and auxiliary agent skills from GitHub repositories. It clones the author's backend template (vibe-config-backend) and the drupal-claude-skills repository. The skill follows security best practices by using a pinned commit hash for the third-party repository dependency.
  • [COMMAND_EXECUTION]: The scripts execute several local commands including git for version control, ddev for container management, and curl for site connectivity verification. The scripts are implemented with safety features such as absolute path resolution, shell variable quoting, and strict error handling (set -euo pipefail).
  • [INDIRECT_PROMPT_INJECTION]: The skill introduces behavioral instructions into the project workspace via the AGENTS.md file and by installing additional agent skills. These instructions influence how the agent interacts with the Drupal site in future sessions.
  • Ingestion points: External content is fetched from Git repositories in scripts/setup-backend.sh and scripts/install-drupal-skills.sh.
  • Boundary markers: The generated AGENTS.md content is wrapped in <!-- BEGIN vibe-config --> and <!-- END vibe-config --> markers to identify the injected instruction block.
  • Capability inventory: The skill can execute shell commands, manage local Docker containers via DDEV, and perform HTTP health checks using curl.
  • Sanitization: The scripts include checks to ensure project folders are valid and located outside of the skill's own directory to prevent unintentional file system operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 20, 2026, 06:49 AM