Development Setup¶
Install¶
--locked is required. The harness measures a pinned TrustSight; a test run
that resolves fresh dependencies is not measuring the same thing.
Run the test suite¶
The suite covers the Judge matrix, validator calibration, DB restore and canary, deduplication, generators, and the self-security gates.
Run the gates individually¶
# Validator calibration
uv run pytest tests/test_validators.py -q
# Self-security gates
uv run pytest tests/test_self_security.py -q
# Secret scan
uv run python scripts/scan_secrets.py
# Lint
uv run ruff check .
Run the shipped campaign¶
Build the docs companions¶
This writes site/llms.txt and site/llms-full.txt from zensical.toml's
navigation. A missing or stale companion fails CI.