ai-repo-safety

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The CLI tool orchestrates multiple security scanners such as Gitleaks, TruffleHog, Bandit, and Ruff by executing them as subprocesses within the repository context. This is the primary intended function of the skill.
  • [EXTERNAL_DOWNLOADS]: The install-tools command allows for the installation of system binaries. It fetches the TruffleHog binary from its official GitHub repository and leverages standard package managers such as Homebrew, WinGet, and npm for other tools. These downloads originate from well-known and trusted services.
  • [DYNAMIC_EXECUTION]: The skill generates and installs project-local shell scripts and Python runners, including pre-push hooks and an agent_hook_runner.py script. These are stored in the project's .git/hooks and .repo-safety/scripts/ directories to facilitate automated security preflights.
  • [INDIRECT_PROMPT_INJECTION]: The github-guard and gitlab-guard components ingestion data from external issue trackers and pull requests. The skill implements proactive mitigations for this attack surface, including secret redaction and pattern-based detection of prompt injection strings to sanitize data before it enters the AI context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 04:03 PM
Security Audit — agent-trust-hub — ai-repo-safety