Target Architecture Merge Readiness (RFC-0007)¶
Status date: 2026-05-28 Reference RFC: docs/rfcs/RFC-0007-OSS-FIRST-TARGET-ARCH-EXECUTION.md Progress baseline: docs/TARGET_ARCH_PROGRESS.md
Executive Status¶
- Overall strategic completion: 75%
- Policy/Tenancy pillar: 95%
- Tenant isolation matrix: passed (25/25)
- same_tenant capability baseline: passed (8/8)
RFC-0007 Contract Readiness¶
1) Scope classification¶
Status: Complete
Evidence:
- docs/TARGET_ARCHITECTURE.md defines pattern transfer vs optional adapter vs operational substrate
- docs/rfcs/RFC-0007-OSS-FIRST-TARGET-ARCH-EXECUTION.md enforces classification requirement
2) Invariant impact matrix¶
Status: Complete (documented), Partial (continuous enforcement)
Evidence:
- docs/TARGET_ARCHITECTURE.md design invariants section
- .github/PULL_REQUEST_TEMPLATE.md architecture-impact sections
Remaining:
- Keep PR discipline consistent in all architecture-impacting changes
3) Backward compatibility and rollback¶
Status: Complete (for current policy/tenancy slice)
Evidence:
- External policy modes off/shadow/enforce with fail-open/fail-closed controls
- Runtime fallback behavior validated in runtime/test_safety.py
4) Validation evidence plan (local + CI alignment)¶
Status: Complete for policy/tenancy slice
Evidence:
- tooling/verify_policy_shadow_mode.py
- tooling/verify_tenant_isolation_matrix.py
- tooling/verify_policy_bundle_lifecycle.py
- tooling/verify_branch_protection_policy.py
- runtime canary integration in .github/workflows/smoke.yml
6) Branch protection governance¶
Status: Partial (in-repo policy/verifier complete, GitHub settings still operational)
Evidence:
- docs/BRANCH_PROTECTION_POLICY.md
- docs/GITHUB_RULESET_RUNBOOK.md
- docs/PRODUCTION_READINESS.md
- tooling/verify_branch_protection_policy.py
- policy-bundle-governance job integration in .github/workflows/ci.yml
Remaining:
- Enforce branch/ruleset settings in GitHub repository configuration (outside repo code).
- Keep release evidence attached per docs/PRODUCTION_READINESS.md (including manual verification when API status is
unverified).
5) Explicit non-goals¶
Status: Complete
Evidence:
- docs/TARGET_ARCHITECTURE.md non-goals and scope guardrails
- docs/rfcs/RFC-0007-OSS-FIRST-TARGET-ARCH-EXECUTION.md non-goals
Slice-Specific Readiness (Policy/Tenancy)¶
Runtime enforcement¶
Status: Complete
Evidence:
- same_tenant enforcement in runtime execution pre-checks
- tenant propagation through neutral API async/resume/replay/fork
- transport propagation in HTTP and MCP paths
Registry adoption baseline¶
Status: Complete
Capabilities currently enforcing same_tenant baseline:
- capabilities/decision.task.delegate.yaml
- capabilities/web.request.send.yaml
- capabilities/email.message.send.yaml
- capabilities/message.notification.send.yaml
- capabilities/agent.plan.execute.yaml
Test and verification¶
Status: Complete
Evidence:
- runtime safety tests for same_tenant
- policy shadow tests for same_tenant internal parity
- channel tenancy probes (http-async/http-resume/http-replay)
- transport tenancy probes (HTTP auth/body resolution + MCP skill.execute)
Open Gaps Before Declaring Full Target Completion¶
- Durability to Temporal-grade semantics (beyond current baseline contract)
- Contracts/artifacts/promotion orchestration with stronger lineage model
- OPA governance depth beyond manifest checks: branch/CI-enforced environment promotion policy
- Wider same_tenant rollout beyond 8-capability expansion
Commit Guidance with Large Dirty Trees¶
When there are many unrelated modified files, use selective commits by explicit pathspec and separate runtime vs registry commits.
Recommended split:
- agent-skills commit: runtime, customer-facing, tooling, tests, docs in this slice
- agent-skill-registry commit: capability YAML contract updates only
- Exclude generated artifacts from both commits unless intentionally updated