stoneburner

Use Stoneburner's inference.env standard and brain utilities to route model evaluation through a selected backend. Use when reading or writing INFERENCE_BACKEND, INFERENCE_URL, INFERENCE_MODEL, INFERENCE_THINK, /opt/agentic/inference.env, atomics.inference, provider_from_target, brain-status, brain-switch, or brain-vllm.

Evaluate models for AI-backed application, workflow, or challenge compatibility with Stoneburner. Use when running atomics qa, adversarial, sweep, provider-test, model promotion, CTF solvability checks, guardrail regression, or classifying models as function-compatible, walkthrough-compatible, too safe, unsafe, or broken.

Create and review Stoneburner target profiles for app-level AI gates and custom inference endpoints. Use when working with profiles/examples, profiles/local, atomics qa --profile, soak --profile, stress --profile, scenario profile workloads, HTTP target profiles, Ollama target profiles, response parsing, or classification rules.