Beltic specification

Beltic issues verifiable credentials that prove who built an agent and how safely it operates. The spec repository contains human-readable docs, machine-readable schemas, and working examples for the two core credentials. A public schema host (for example, schema.beltic.com) is not live yet, so tooling references relative repo paths. Schemas themselves are not hosted in this site—fetch them from the repository when you validate or integrate.

What lives in beltic-spec/

• Overview and flow: docs/overview.md, docs/quickstart.md, docs/integration-guide.md, docs/merchant-faq.md
• Core specs: docs/developer-credential-v1.md, docs/agent-credential-v1.md, docs/agent-manifest-spec-v1.md
• Mapping and evaluation: docs/nist-mapping-v1.md, docs/evaluation-metrics-v1.md, docs/metrics-roadmap.md
• Validation and usability: docs/validation-guide.md, docs/v1-usability-validation.md
• JSON Schemas (kept in-repo): schemas/developer/v1/developer-credential-v1.schema.json, schemas/agent/v1/agent-credential-v1.schema.json
• Examples and templates: examples/developer-example-v1.md, examples/agent-example-v1.md, JSON fixtures under examples/developer/v1/tests/ and examples/agent/v1/tests/
• Progress and contributing: progress.md, docs/contributing-spec.md

DeveloperCredential v1 (identity root)

• Purpose: establish the legal entity (or individual) that stands behind every agent credential.
• Field groups: core identity, contact, tax/registration, risk and compliance, ownership and control, verification metadata, cryptographic identity.
• Assurance tags: every field notes self_attested, beltic_verified, or third_party_verified expectations. Production issuances expect Beltic-verified as the default.
• Schema path: schemas/developer/v1/developer-credential-v1.schema.json.

AgentCredential v1 (agent binding)

• Purpose: certify a specific agent's identity, capabilities, safety posture, and operational metadata; always linked to a DeveloperCredential.
• Field groups: agent identity and provenance, technical profile, tools and actions, data handling and privacy, safety and robustness metrics, operations and lifecycle, risk/assurance metadata, cryptographic identity.
• Schema path: schemas/agent/v1/agent-credential-v1.schema.json.

Manifests and integration flow

• Agent manifest spec: docs/agent-manifest-spec-v1.md covers the developer-authored manifest used during AgentCredential issuance.
• Integration guide: docs/integration-guide.md shows how verifiers and merchants consume credentials, handle policy, and track revocation.
• Merchant FAQ: docs/merchant-faq.md answers onboarding and interpretation questions.

Examples and templates

• Human-readable examples: examples/developer-example-v1.md, examples/agent-example-v1.md.
• JSON fixtures:
  • Developer templates: examples/developer/v1/tests/valid-individual-*, valid-organization-*, plus invalid sets.
  • Agent templates: examples/agent/v1/tests/valid-* and invalid variants.
• Quickstart templates: docs/quickstart.md points to starter files and shows validation commands.

Validation commands (from beltic-spec)

Install dependencies once, then run validators:

npm install
npm run validate:all

Focused checks:

npm run validate:agent
npm run validate:developer
npm run test:conditional-rules   # 27 developer rules

Python validator:

pip install jsonschema
python3 scripts/validate_all.py

See docs/validation-guide.md for AJV/jsonschema setup, runtime checks (dates and freshness), CI integration, and troubleshooting.

Versioning and assurance model

• v1 artifacts live beside future versions (v2, v3, ...) so integrators can pin compatibility; each schema carries a schemaVersion.
• Assurance levels are carried per field to signal whether data was self-attested, Beltic-verified, or verified by an independent party.
• AgentCredentials and DeveloperCredentials are designed to be portable across ecosystems: relying parties choose which issuers or assurance tiers to accept.

Evaluation and mapping references

• docs/evaluation-metrics-v1.md: attack success rate, robustness scores, required metadata.
• docs/metrics-roadmap.md: upcoming scoring dimensions.
• docs/nist-mapping-v1.md: mapping to NIST AI RMF (GOVERN/MAP/MEASURE/MANAGE).
• docs/v1-usability-validation.md: feedback and adjustments from early adopters.