System philosophy
SAIRP separates intent, authority, and execution. The three are distinct objects with distinct lifecycles, and the protocol's correctness rests on never collapsing them.
Execution must consume already-verified authority, not infer it. A runtime that derives permission from prose, diagnostics, or prior outputs has merged authority into execution — a category violation that the protocol exists to prevent.
Classification of this document: Operational Entry, non-authoritative for execution permission. It orients readers and points at the canonical artifacts; it does not grant any authority on its own.
Core docs
The protocol is specified across a small set of canonical artifacts. Read them in this order to walk from architecture down through governance into the runtime phases.
docs/00_ARCHITECTURE_PHASE_BOUNDARIES.md— phase boundary definitionsdocs/01_governance/— governance law set (artifact roles, authority lattice, version manifest)docs/runtime/04_negotiation/00_NTA_USAGE.md— Phase 4 negotiationdocs/runtime/05_agreement/— Phase 5 agreement and ACK envelopedocs/runtime/06_admission/— Phase 6 admission and 6.5 handshakedocs/runtime/08_operations/SAIRP_meetLab_Operations_Guide.md— operator surface
For a single-page walk through the runtime phases, see the lifecycle walkthrough.
Build & operations entry
Canonical build entry
make tier0 tier1 tier2 ORCHESTRATED=1
Governed artifacts are promoted under artifacts/ and runtime state under runtime/. Anything outside those roots is not part of the governed surface and cannot be cited as authority.
Audit reminder
For canonical HALT semantics, reference docs/runtime/07_execution/halt_codes.md.