Policy

Substrate Scope Policy

Which execution substrates may appear in the Validation Lab and under what admission criteria.

Governance Boundary

Non-certification · Evidence-based verdicts only · No endorsement · No execution hosting · Full statement →

1. Key Principle

  • •This policy does not define MPLP protocol requirements.
  • •This policy does not certify or endorse any framework.
  • •The Lab adjudicates evidence, not execution.

2. Scope Tiers

Tier-0Canonical Target List

The minimal mainstream set required to credibly claim cross-framework adjudication.

Execution Frameworks (4)

  • • LangGraph
  • • AutoGen
  • • Semantic Kernel
  • • LangChain

Protocol Rails (2)

  • • A2A
  • • MCP

📊 SSOT: data/curated-runs/substrate-index.yaml

Tier-1Extended / Rolling

Frameworks with meaningful adoption (e.g., LlamaIndex, Haystack)

Tier-2Ecosystem / Long-tail

Niche, experimental, or closed-source substrates

3. Claim Levels (Admissibility)

Each substrate evidence pack carries a claim level indicating its reproducibility:

LevelDefinition
DeclaredStatic or mocked producer; evidence present but not reproduced
ReproducedDeterministic reproduction demonstrated (run-twice hash match)
Cross-VerifiedEquivalence evidence across paradigms under the same ruleset

4. Admission Criteria

For an evidence pack to be admitted (ADMISSIBLE), it must pass these structural gates:

  • MAN-001All required manifest fields present (pack_version, protocol_version, etc.)
  • INT-001SHA256 sums verification passes for all listed files
  • INT-002Pack root hash matches computed hash
  • INT-003All pack files covered by SHA256 sums (no uncovered files)
  • VER-*Protocol version binding and schemas bundle hash verification

See Intake Policy for status taxonomy (REGISTERED, ADJUDICATED, NOT_ADMISSIBLE, etc.)

Traceability

This page is a public projection of the internal governance document. Authoritative source: SUBSTRATE_SCOPE_POLICY.md ↗

← Adjudication CoverageIntake Policy →