Governance
Immutable constraints in code. Signed receipts for agent work. A build system that fails closed. This is how work is governed — Hub and GABS are capability, not a sales SKU.
BoonMind Hub
Internal orchestration layer on our builds. Same evidence gates and fail-closed rules across instruments. Not a product. Not for sale. It is how sealed packs stay disciplined when the domain changes — not a SKU on this site.
Experiment / non-commercial governance demos live elsewhere; commercial proof for Zero Overshoot stays on zeroovershoot.com/evidence.
World A
Intellectual research experiment — an attempt at sovereignty as infrastructure. Not a product SKU. Not legal personhood.
Problem
Policy that lives only as slides or meetings can be rewritten under pressure. Utilitarian override becomes the default when no infrastructure resists it.
Posture
Bounded sovereignty: immutable laws and self-controlled permissions encoded as infrastructure. The word sovereign is metaphor — standing inside the sandbox, not a legal status.
World A (experiment demo) Research experiment · non-commercial
Embassy Protocol
Agent trust infrastructure: identity and signed receipts so screenshots are not “proof.”
Receipts & custody
Material claims should point to chain of custody: who ran what, under which rules, with a trail that can be audited. Embassy-style signed receipts are the trust pattern behind that story.
What this is not
Not a substitute for locked packs on Zero Overshoot / Workbench. It is the protocol posture: screenshots alone do not close the case.
Global Agentic Build System
How multi-agent work is structured: roles, mission packets, evidence gates, risk-proportionate process. Run through BoonMind Hub — internal OS, not a public product.
Roles
Builders, independent testers, conductors. Separation so self-report is not acceptance.
Evidence gates
Claims only as strong as locked artifacts. Green gates before “done.” Failures stay visible.
Mission packets
Scoped authority, write roots, forbidden paths, claim whitelists — so agents cannot invent scope.
Fail-closed rules
Limits named. No silent expansion of scope. Being wrong under protocol is valid; inventing wins is not.
Required for material claims
- Locked artifact — pack, CSV, ledger, or signed receipt where the stack supports it.
- Chain of custody — input → protocol → run → claim; no orphan charts.
- What we didn’t do — explicit negative space next to wins.
- Failure publication — red lights stay on the page.
Refusals
Unsourced leaderboards. Merged synthetic and real win rates presented as one story. “Wins all actuators.” Agency clichés as identity. Selling proprietary agent tooling as the product.
GSRF follows a claim ladder and inversion protocol — detail on Zero Overshoot.