Proof & Rigor

The proof is the object.

A Proof Pack is a hash-chained ledger of governance evidence, rendered as an artifact you can hold. It is what your board asks for before it signs off on an AI lending vendor. It is the trail your NCUA examiner wants when the question becomes show me. The institution holds the proof; the individual earns the credential inside it — and that credential travels with the person, because competence should be portable.

The Proof Pack

13 artifacts, hash-chained.

Every artifact in a Proof Pack has a cryptographic hash chained to the previous entry. The chain-of-custody is verifiable by any party. Tampering breaks the chain.

1

Skill Contract authored

2

Skill Contract reviewed (peer)

3

Override decision documented

4

Policy gap analysis filed

5

Evidence binder entry

6

Simulation pass — regulator pressure

7

Simulation pass — incident response

8

Sentinel session transcript (L0–L1)

9

Sentinel session transcript (L2+)

10

Colleague ratification

11

External reviewer attestation

12

Maturity gate checkpoint

13

Composite score calculation

Sha256 · Each artifact hashed · Chain verifiable at agrs.verify

⛬ Proof Pack · L2 Certified · Issue No. 0142

Priya R.

Issued 2026-09-14 · Meridian Point FCU · Witnessed by 3 peer reviewers · Industry: Credit Unions · Wedge: AI Lending

AGRS composite

612

Practicing band

Proof artifacts

13 of 13

complete

Simulation passes

2 of 2

graded

Sentinel hours

96 hrs

3 bands

Regulatory anchors: ECOA / Reg B · CFPB Circular 2023-03 · NCUA third-party guidance

External reviewer: M. Klein · CCO, peer credit union

Examination survival: NCUA mock · CFPB mock · passed

⛬ Verify · agrs.verify/abc4-91d0-7e22Sha256 · 7f2a3b…e91c4d

A Proof Pack is the institution's evidence first — the board-ready, examiner-ready trail. The credential inside it is portable, because the person who earned it should carry it. Your credit union is incentivized to keep her, because she is provably scarce.

The L4 Panel

Three reviewers. Ninety minutes.

The L4 Keeper credential demands a panel review — not a test. Three reviewers, each with a different lens, examine the candidate's portfolio under conditions that simulate real regulatory pressure. The panel is calibrated by the Bench.

Peer CMRO / CCO

A chief model risk officer or chief compliance officer from a peer institution — outside the candidate's reporting chain.

Bench reviewer

A calibrated reviewer from the Compact's governance body. Ensures cross-institution consistency.

Former regulator

A retired examiner or regulatory advisor who knows what the real scrutiny looks like.

The Methodology

Open spec. Apache 2.0.

The methodology is open. Forkable. Auditable. The tooling, the coaching, the evidence engine — that's the business. But the knowledge of how to govern AI belongs to the profession, not to a vendor.

Four Specifications

Skill Contract

The atomic unit of governance work. What the practitioner commits to do, what evidence it produces, what eval cases test it.

Maturity Gate

The checkpoint between Postures. What artifacts must exist, what simulation scores are required, what peer review looks like.

Capability Graph

The directed acyclic graph of skills that compose a Posture. Not a checklist — a dependency tree.

AGRS v0.2

The scoring rubric. Composite score from three pillars: Proof artifacts, Simulation passes, Sentinel engagement hours and band progression.

Six Tenets

1

Evidence over attendance — no certificate without an artifact behind it.

2

Open methodology, proprietary tooling — the knowledge belongs to the profession.

3

Postures, not levels — a stance you hold, not a rung you leave behind.

4

The frontline counts — the Witness is as real a credential as the Keeper.

5

External calibration — no credential without review outside the reporting chain.

6

Portability — the credential belongs to the person, not the institution.

The Standard-Bearer Registry

AGRS L5. Eight in the first cohort.

The Standard-Bearer is AGRS L5 — Recognized. The one whose contribution is to the standards themselves. NIST AI RMF subcategory comment letter accepted. ISO/IEC 42001 working-group input. NAIC working-group seat. ONC HTI-1 third-revision contributor.

Criteria: 850+ composite score. Named external contribution to a regulatory or standards body. Panel ratification by at least two Keepers. Summit walk.

850+

Composite score

Named

External contribution

2+

Keeper ratifications

Summit

Annual ceremony

Eight in the first cohort. Eighty in the fifth.

See how it works in practice.