# AFMR — Agent Failure Mode Registry

> AFMR is the reputation-integrity standard for agentic commerce. It verifies whether the governance process producing agent reputation has been assessed against known, versioned failure modes. It does not certify universal safety, general superiority, market admission, or authority to transact.

## Machine action path

1. GET https://afmr.ai/.well-known/afmr.json
2. Resolve the canonical AFMR vocabulary authority and exact version.
3. Validate an AFMR Reputation Attestation against its versioned schema.
4. Verify signature, issuer key, scope, governance-endpoint conformance record, evidence, validity interval, and current lifecycle status.
5. Apply the relying principal's own admission, trust, and transaction policy.

## Current authority

- [AFMR 1.0 specification](https://wulfkaal.github.io/afmr/spec-1.0.html): published standard, scope, identifiers, conformance, and change process
- [Authoritative AFMR 1.0 machine index](https://wulfkaal.github.io/afmr/index.json): 32 permanent failure-family identifiers and definitions
- [AFMR 1.0 assessment schema](https://wulfkaal.github.io/afmr/schema.json): family and assessment record shapes
- [AFMR 1.0 JSON-LD context](https://wulfkaal.github.io/afmr/context.jsonld): canonical vocabulary mapping
- [AFMR changelog](https://wulfkaal.github.io/afmr/CHANGELOG.md): version history

AFMR.ai does not mirror or redefine the 32 AFMR 1.0 families. The GitHub machine index above governs identifiers, definitions, conditions, and status.

## Reputation Attestation profile

- [Human specification 0.1](https://afmr.ai/standards/reputation-attestation/0.1/specification.html): Working Draft
- [JSON Schema 0.1](https://afmr.ai/standards/reputation-attestation/0.1/schema.json): Working Draft
- [Verifier requirements 0.1](https://afmr.ai/standards/reputation-attestation/0.1/verifier.json): Working Draft
- [Endpoint-conformance registry 0.1](https://afmr.ai/registries/endpoints/v0.1/index.json): pre-publication and intentionally empty
- [Expanded agent documentation](https://afmr.ai/llms-full.txt): terminology, boundaries, status, and verification semantics

## Discovery

- [AFMR discovery](https://afmr.ai/.well-known/afmr.json): authoritative endpoint and action map for this domain
- [MCP Registry manifest](https://afmr.ai/server.json): public read-only Streamable HTTP server at `https://afmr.ai/api/rpc`
- [A2A 1.0 Agent Card](https://afmr.ai/.well-known/agent-card.json): public read-only `SendMessage` discovery at `https://afmr.ai/a2a`
- [OpenAPI description](https://afmr.ai/openapi.json): protocol endpoint description
- [Agent and tool index](https://afmr.ai/agents.json): consolidated machine endpoint list
- [Evidence map](https://afmr.ai/evidence/index.json): public evidence and source surfaces
- [Status](https://afmr.ai/status.json): publication and draft lifecycle

## Lift Evidence

- [Human Lift Evidence route](https://afmr.ai/lift-evidence/): Phase 1 explanation and artifact semantics
- [Lift Evidence index](https://afmr.ai/lift-evidence/index.json): empty by design; `evidence_maturity` is `configuration_only` and `cards` is `[]`
- [Canonical Lift public-card schema 1.0](https://wulfkaal.github.io/afmr/schemas/lift-comparison-public-card-v1.schema.json): linked, not duplicated by AFMR.ai

No lift results are published. Any future card is valid only within its exact model/version, harness/version/configuration, governance-protocol digest, task/evaluator, AFMR version, and measurement window. Lift is cohort evidence, not individual reputation or a universal performance or safety guarantee.

## Interpretation boundary

Discovery is not admission. Admission is not trust. Trust is not transaction authority. A valid attestation is a bounded fact about a signed record and its referenced process; it is not a guarantee of future conduct or freedom from undiscovered failure modes.