Your AI output isn't production-grade until
SQG says it is.
36 anti-slop patterns (A-001 through A-036). 14 quality dimensions. One executable gate. Built by the Saffron AI Group — setting the standard the rest of AI will be measured against.
Three commands. One standard.
SQG turns the Saffron Quality Benchmark Canon — 752 lines of engineering philosophy — into executable, verifiable checks. No AI. No hallucination. No excuses.
sqg scan
Scans your entire project against all 36 anti-slop patterns (A-001 through A-036). Catches silent failures, placeholder logic, hardcoded values, timezone neglect, empty tests, copy-paste reuse, and 30 more classes. Each check is deterministic.
sqg audit
Scores your deliverable across the 14 Saffron Quality Dimensions. Maps real tool output (ruff, mypy, bandit, the anti-slop scanner) through weighted rubrics. Produces an Appendix B-certified audit report.
sqg gate
The unified quality pipeline. Runs anti-slop scan + dimension audit + doctrine integrity. One command, one exit code. 0 = all gates pass. Wires into CI/CD, pre-commit hooks, and deploy pipelines.
Enterprise Titanium Grade
SQG enforces the same standard that powers every Saffron deliverable — zero bare excepts, full type annotations, verified inputs, immutable self-healing.
10+ Tools
ruff, mypy, bandit, semgrep, mutmut, pytest-cov, pa11y, vale, eslint, prettier, lighthouse — all mapped to quality dimensions.
CI/CD Ready
JSON output, exit codes, strict mode. Wires into GitHub Actions, GitLab CI, Vercel. Pre-commit hook in one line.
No Cloud
Runs entirely on your machine. Zero data leaves your network. The standard is executable, not a service call.
Extensible
Every check is a Python function. Add custom patterns, adjust severity thresholds, YAML config included.
Install the gate.
One command. All 36 patterns. Zero excuses.