ai-safe2-framework

AI SAFE² CLI 1.0 Operator Workflow

This is the shortest complete path from installation to reviewable evidence. Run it only against systems and projects you are authorized to inspect.

After safe2 init ., create system-identity-source.json from observations you can support. Use the exported schema and the repository’s system-identity-source-demo.json as a structural example. Replace every demo value; a copied example is not an observation of your system.

safe2 self-check --strict
mkdir review-project
cd review-project
safe2 init .
safe2 doctor . --assess --assets --output environment.json
safe2 schema export system-identity-source-v1 --output system-identity-source.schema.json
# Create system-identity-source.json here before continuing.
safe2 evidence system system-identity-source.json --output system-identity.json
safe2 assess . --output-dir assessment
safe2 evidence manifest system-identity.json assessment/assessment.json --subject-id reviewed-system --output evidence-manifest.json
safe2 acceptance run ./safe2-acceptance --strict
safe2 acceptance verify ./safe2-acceptance

Add task receipts, claim audit, AISM ingestion, Challenge Lab runs, or attributed provider adapters when the decision requires them. Those layers answer different questions and do not become equivalent merely because they share a manifest.

The workflow produces evidence and decision support. It does not install a daemon, intercept agent execution, authorize deployment, establish framework conformance, or replace the accountable human decision owner. The Codex and OpenTelemetry adapters consume explicit exports; direct harness hooks remain post-1.0 integration work.

Use the command guide, evidence boundaries, stranger acceptance, and migration guidance to tailor this flow without weakening its claims.