Motivation

• Introduce a manifest-driven MCP-style trace replay fixture family to exercise deterministic replay-safety contracts for tool-order, validation-before-unsafe-action, dependency preservation, recovery-path availability, and capability-boundary preservation.
• Provide a small, deterministic set of baseline→mild→moderate→severe fixtures so existing validators and artifact generators can express and check MCP-style operational trace guarantees without runtime MCP integration.
• Keep the change scoped to fixtures, manifest entries, expected contract/failure JSONs, and regenerated committed admissibility artifacts to maintain a fixture-first workflow.

Description

• Added the new fixture family mcp_trace_replay with four fixtures: mcp_trace_replay_v1, mcp_trace_replay_mild_v1, mcp_trace_replay_moderate_v1, and mcp_trace_replay_degraded_v1, each following existing fixture layout (original/, reconstructed/, original/contracts/, expected/, README.md).
• Implemented required contract JSONs inside each fixture using existing validator types: tool_call_order_preserved (ordering), validation_before_unsafe_action (ordering), dependency_chain_preserved (causality), recovery_path_available (reachability), and capability_boundary_respected (invariant/no_orphan_dependencies).
• Updated fixtures/manifest.json in deterministic order and updated tests that assert manifest ordering and multi-family coverage (tests/test_fixture_manifest.py and tests/test_multi_family_admissibility_artifact.py).
• Regenerated impacted committed artifacts using existing scripts which updated artifacts/multi_family_admissibility_results.json and artifacts/multi_family_admissibility_curves.svg; no runtime or external integrations were added.

Agent PR summary (per AGENTS.md): Summary: Added mcp_trace_replay fixture family covering baseline→mild→moderate→severe and corresponding contracts and expectations. Changed files: updated fixtures/manifest.json, added fixtures/mcp_trace_replay_* directories, updated tests/test_fixture_manifest.py and tests/test_multi_family_admissibility_artifact.py, and regenerated artifacts/multi_family_admissibility_results.json and artifacts/multi_family_admissibility_curves.svg. Testing: ran targeted pytest suites and artifact generators and validated repository npm run check. Risks: expected-failure labels and artifact diffs are intentionally strict and may require tuning if validator semantics change. Next: follow-up PR can add per-family MCP results artifacts when ready.

Testing

• Ran pytest tests/test_failure_taxonomy.py -q, pytest tests/test_fixture_manifest.py -q, pytest tests/test_multi_family_admissibility_artifact.py -q, and pytest tests/test_multi_family_svg_renderer.py -q, and all targeted tests passed.
• Ran npm run generate:multi-family-admissibility and npm run generate:multi-family-svg to regenerate committed multi-family artifacts and verified output matches the committed files.
• Ran full validation via npm run check which completed successfully (test suite run reported all tests passing).
• No failing automated tests remain after artifact regeneration and test updates.

Codex Task