Creates
A config, presets lock, project-local skills, process guidance, examples, and an onboarding packet across 19 recorded file actions.
Agent-ready onboarding
Give the agent a baseline without teaching it to guess.
One onboarding command creates or preserves the broad Assura baseline, installs project-local guidance, verifies the result, and leaves project-specific questions for a human decision.
Assura agent onboardingProject node confidence=highAgent codex confidence=explicitPolicy agentic-core + structure-health -> $agent-entrypoint (applied)Host codex generated=true activated=true structurally_verified=true conflicted=falseHost status post_activation=structurally_verified doctor=pass host_approval=unavailable follow_up=Review the generated project hooks, then trust the project and approve its hooks in Codex.Lifecycle agent-working-loop=nudge pre-commit-warning=warn pre-push-or-ci-gate=gateVerified structure_config=pass onboarding_packet=passReview needs-review blocking=0 advisory=0 inactive_signals=5Deferred project_specialization, content_modelsNext Specialize from repository evidencePacket .assura/onboarding/agent-next.mdPreserves
Existing configuration and AGENTS.md guidance are merged or retained rather than silently replaced.
Does not guess
Language, layout, naming, hooks, content models, and domain rules remain explicit human decisions.
Active versus inactive
The onboarding result names verified setup separately from content models and specialization that are still inactive.
Harness-aware, not harness-dependent
Supported harnesses can receive integration bundles; generic environments still get AGENTS.md and manual-shell guidance.
A concrete handoff
The generated agent-next document tells the agent what to read, what to ask, and which broad checks are ready.
One lifecycle, separate pressure
Advisory working-loop feedback, pre-commit warnings, and blocking pre-push or CI policy are configured as distinct stages.
Continue