Add Wargames adversarial scenario contract

[mdheller]

Summary

Adds the first SCOPE-D Wargames adversarial scenario composition contract.

This is the executable JSON-schema counterpart to the ProCybernetica scenario-boundary doctrine. The new contract makes AdversarialScenario a governing composition object over existing Wargames artifacts rather than another ATT&CK coverage cell.

Adds

• config/schemas/wargames-adversarial-scenario.schema.json
• valid scenario example:
  • examples/scope-d/wargames/wargames-adversarial-scenario.example.json
• negative fixtures:
  • examples/scope-d/wargames/negative-fixtures/adversarial-scenario-attack-only.invalid.json
  • examples/scope-d/wargames/negative-fixtures/adversarial-scenario-memory-writeback.invalid.json
• validator wiring in scripts/validate-wargames-contracts.js

Contract semantics

The scenario schema binds:

• objective and target topology
• channel substrates and interface crossings
• authority envelope and safety boundaries
• capability exposure, all fail-closed in examples
• human and machine interpretation risks
• boundary events, evidence refs, ATT&CK coverage claims, controls, proofs, runtime decision receipts
• memory effects and review-only writeback posture
• claim-promotion state
• consequence model
• abstention rules
• counterfactual branches
• semantic non-claims and redaction state

Safety boundary

This PR adds no runtime, no live collector, no engagement execution, no command execution, no payload delivery, no credential access, no memory writeback, no public report publication, and no downstream activation authority.

The schema deliberately holds these capability exposures at false for Wargames examples.

Validator behavior

The Wargames validator now requires the valid scenario fixture to pass and rejects invalid fixtures when:

• ATT&CK/local coverage is treated as complete without scenario semantics or runtime receipt grounding.
• Scenario learning attempts to bypass Memory Mesh review-only proposal governance.

Next tranche

P0.3: harden SCOPE-D semantic scenario gates further with additional negative fixtures for synthetic promotion and summary-only/model-generated finding promotion.