Protect the workflow. Don't stop the workflow.
Public, inspectable artifacts from AI product work: synthetic eval harnesses, workflow gates, case studies, and a proof pack focused on evaluation, refusal/uncertainty states, and release judgment for constrained domains. Gate mechanics defer to agent-proof-kit. Synthetic, not production data.
Codex skill that keeps project files, chat context, operator instructions, working hypotheses, and model assumptions in separate cognitive layers, with a compact frame and optional PreToolUse hook.
A context-preserving Git/GitHub safety skill for AI coding agents.
Deterministic, network-free release gates for AI-agent runs. Validates a JSON run/policy contract, blocks high-risk actions without explicit approval, fails closed on unknown actions, scans the public surface, exports SARIF, and runs as a GitHub Action. Tested against synthetic fixtures; CI green.
A cognitive layer for Claude that tracks epistemic weight during complex working sessions: what is still a hypothesis, what has been decided, and what Claude generated but you have not validated yet.
Profile README for Guillaume Vele — AI product builder (evals, workflow safety, on-device voice). Co-founder & digital director at Nuance Paris. Open Q8_0 quantization PR to TrevorS/voxtral-mini-realtime-rs.
Add a description, image, and links to the workflow-safety topic page so that developers can more easily learn about it.
To associate your repository with the workflow-safety topic, visit your repo's landing page and select "manage topics."