ai-coding-agent-guardrails
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Downloads the
git-secretstool from the official Amazon Web Services (awslabs) GitHub repository. - [EXTERNAL_DOWNLOADS]: References and integrates trusted security tools and images, such as Semgrep, TruffleHog, and official Ubuntu Docker images.
- [COMMAND_EXECUTION]: Provides scripts for system-level configuration including user creation (
useradd), filesystem access control (setfacl), and firewall rule configuration (iptables). - [COMMAND_EXECUTION]: Uses standard development and CLI utilities such as
npm,pytest,git, and the GitHub CLI to manage CI/CD pipelines and auditing. - [PROMPT_INJECTION]: Processes AI-generated content (code, branch names, and PR descriptions), which serves as an indirect prompt injection surface.
- Ingestion points:
scan_agent_output.py(file content reading) and.github/workflows/agent-pr-review.yaml(PR metadata and body). - Boundary markers: Recommends defining constraints in
CLAUDE.md,.cursorrules, andcodex.mdfiles. - Capability inventory:
setup-agent-workspace.sh(system permissions),agent-network-controls.sh(egress filtering), andrun-agent-sandbox.sh(container execution). - Sanitization: Employs regular expressions to identify and block potential secrets, although it does not perform content sanitization or escaping.
Audit Metadata