# Deterministic agent evaluation gates

This dependency-free Deno fixture compares a synthetic baseline and candidate agent
release against case-specific invariants. It records newly introduced failures and
blocks only the severities declared by the fixture configuration.

The fixture does not call a model, tool, external API, or paid service. Its task names,
policies, costs, releases, and results are synthetic. It does not measure model
quality, real latency, production cost, or judge calibration.

## Run

```bash
deno fmt --check *.ts
deno check agent_gate.ts fixture.ts run_fixture.ts render_manifest.ts \
  agent_gate_test.ts evidence_artifact_test.ts
deno lint *.ts
deno test --allow-read agent_gate_test.ts evidence_artifact_test.ts
deno run run_fixture.ts
deno run --allow-read --allow-write render_manifest.ts \
  agent-evaluation-manifest-2026-08-29.json \
  agent-evaluation-gate-2026-08-29.svg
```

The saved manifest is a deliberately failing candidate release used to prove that the
gate records every reason and blocks only the configured severities. Code is licensed
under MIT. The checklist, manifest explanation, and diagram are licensed under CC BY
4.0.