Accounting Runtime Promotion Design

Goal

Promote accounting from activation_status: eval_ready to activation_status: runtime_enabled only after the merged accounting fixture pack, subject gate, and full-cycle workflow harness evidence support the activation.

This is a narrow activation slice. It should make accounting available to the adaptive runtime for clear accounting research requests while preserving the core router rule that method-only evidence borrows accounting lenses instead of switching the primary subject.

Current Baseline

On dev as of July 5, 2026:

• The full-cycle workflow harness and manuscript-first journal-fit recommender from the July 4 plan are present.
• The accounting eval-ready fixture pack and manifest-backed signals are merged.
• qiongli_lifecycle_plan and qiongli_journal_fit_recommend are exposed by the Python MCP handler layer.
• tooling/scripts/run_full_cycle_workflow_harness.py runs the deterministic clean empirical fixture successfully.
• accounting has activation_status: eval_ready.

Current gate evidence:

accounting --gate eval-ready:
  activation_status: eval_ready
  eligible_for_eval_ready: true
  blocking_failures: []
  metrics:
    decision_accuracy: 1.0
    primary_subject_accuracy: 1.0
    suggest_subject_precision: 1.0
    near_miss_false_positives: 0
    forbidden_subject_accuracy: 1.0
    method_lens_accuracy: 1.0
    all_case_checks_passed: 1.0

accounting --gate runtime-enabled:
  activation_status: eval_ready
  eligible_for_runtime_enabled: false
  blocking_failures:
    - activation_status is eval_ready
  metrics:
    decision_accuracy: 1.0
    primary_subject_accuracy: 1.0
    suggest_subject_precision: 1.0
    near_miss_false_positives: 0
    forbidden_subject_accuracy: 1.0
    method_lens_accuracy: 1.0
    all_case_checks_passed: 1.0

The runtime gate is blocked only because the manifest has not been promoted. The fixture metrics already satisfy the accounting activation thresholds.

Decision

Run a guarded accounting activation slice now, before adding business, political economy, geoeconomics, or the economics-accounting bridge.

Why this order:

• The full-cycle harness is already merged, so repeating that work would waste the next roadmap slot.
• Accounting is already eval-ready and has zero measured false positives in the current fixture pack.
• The activation diff can be small and reversible.
• Activating one adjacent subject tests the subject-gate contract before broader, less bounded subjects are added.

Scope

In scope:

• Add one extra accounting method-only auto-mode fixture to guard against runtime over-activation after promotion.
• Promote content/subjects/accounting/runtime-subject.yaml to activation_status: runtime_enabled.
• Update tests that currently assert accounting is eval-ready.
• Update CLI, release, and roadmap docs so accounting runtime activation is described accurately.
• Re-run subject gate checks for accounting, finance, and economics.
• Re-run deterministic full-cycle harness checks.

Out of scope:

• Do not activate business, political economy, geoeconomics, or economics-accounting.
• Do not change provider configuration, Zotero lookup behavior, literature search defaults, or full-text retrieval.
• Do not add network-dependent tests.
• Do not launch local agents in default verification.
• Do not broaden accounting signals unless a fixture proves the need.

Runtime Safety Rules

Accounting can become the primary subject only when evidence includes at least two independent accounting dimensions, such as:

• method plus data/outcome,
• method plus theory/construct,
• venue plus accounting construct,
• data/outcome plus accounting construct.

Accounting must not become the primary subject when the evidence is limited to:

• accrual quality or discretionary accruals as a single method-only phrase,
• generic bookkeeping or budget reporting,
• generic "accounting for heterogeneity" language,
• operational financial reporting dashboards,
• management forecast staffing or project operations language.

Method-only accounting evidence may still borrow:

• accrual-quality,
• construct-proxy-audit.

Borrowed accounting lenses should use resource_level: method_pack_only/method-level activation and must not load full accounting subject resources unless accounting is the active or suggested subject.

Required Fixture Addition

Add a new auto-mode method-only guard:

tests/fixtures/subject_router_eval/accounting/method_only_auto_accrual_controls.json

Expected behavior after accounting is runtime-enabled:

• Decision: borrow_lens.
• Primary subject: auto.
• Suggested subjects: none.
• Forbidden subjects: accounting.
• Method lenses: accrual-quality.

This fixture is intentionally different from the existing locked-finance method-only case. It proves that runtime-enabled accounting still does not activate from a single accounting method signal in a neutral auto project.

Expected Code Changes

Accounting manifest

Change:

activation_status: eval_ready

to:

activation_status: runtime_enabled

No signal weights should change unless the new method-only guard fails.

Subject contract tests

Update repository contract expectations:

• subject_activation_status("accounting", contracts) becomes runtime_enabled.
• The accounting manifest test should be renamed from eval-ready wording to runtime-enabled wording.
• Accounting must still declare signal groups, method lenses, evaluation pack, near-miss policy, and gate metrics.

Subject router gate tests

Update real accounting gate tests:

• --subject accounting --gate runtime-enabled must pass.
• The old real-fixture assertion that accounting runtime gate blocks on activation_status is eval_ready should be replaced with a patched unit test that still proves an eval-ready contract cannot pass the runtime gate.
• Existing patched eval-ready gate tests should remain, because future subject candidates still use that path.

Subject refinement tests

Update runtime behavior expectations:

• Clear accounting evidence in auto mode can suggest accounting.
• Method-only accounting evidence in auto mode borrows the accounting lens without suggesting accounting.
• Confirmed accounting projects can load accounting resources now that accounting is runtime-enabled.
• Locked finance/economics projects still keep the lock and may only borrow accounting method lenses.

Verification

Required focused checks:

uv run python -m pytest \
  tests/test_subject_contracts.py \
  tests/test_subject_router_eval.py \
  tests/test_subject_refinement.py \
  -q

Required gate checks:

uv run python tooling/scripts/evaluate_subject_router.py \
  --subject accounting \
  --gate runtime-enabled \
  --json

uv run python tooling/scripts/evaluate_subject_router.py \
  --subject finance \
  --gate runtime-enabled \
  --json

uv run python tooling/scripts/evaluate_subject_router.py \
  --subject economics \
  --gate runtime-enabled \
  --json

Required regression checks:

uv run python tooling/scripts/evaluate_subject_router.py --json

uv run python tooling/scripts/run_full_cycle_workflow_harness.py \
  --fixture tests/fixtures/full_cycle_harness/clean_empirical \
  --json-report /tmp/qiongli-full-cycle-harness.json

uv run python -m pytest \
  tests/test_full_cycle_harness_script.py \
  tests/test_lifecycle_harness.py \
  tests/test_journal_fit.py \
  tests/test_mcp_tool_handlers.py \
  -q

git diff --check

Acceptance Criteria

• Accounting runtime-enabled gate exits 0.
• Accounting fixture metrics still meet all manifest thresholds.
• Accounting near-miss false positives remain 0.
• The new auto-mode method-only fixture does not suggest accounting.
• Finance and economics runtime-enabled gates still pass.
• Full-cycle workflow harness still passes the clean empirical fixture.
• Docs no longer describe full-cycle harness as the next unimplemented priority.
• Business, political economy, geoeconomics, and economics-accounting remain unactivated.

Risks

• Over-activation: accounting may become suggested from weak method-only language. Mitigation: add the auto-mode method-only fixture before promotion.
• Adjacent-subject drift: accounting could override finance/economics when it should only contribute a measurement lens. Mitigation: keep locked and mixed subject fixtures in the runtime gate.
• Resource loading mismatch: confirmed accounting may begin loading subject resources that were previously withheld. Mitigation: update tests to assert the exact resource activation plan.
• Roadmap drift: full-cycle status may remain stale in docs. Mitigation: update the roadmap in the same PR as the activation plan or implementation.

Rollback

If the activation causes routing drift:

1. Set content/subjects/accounting/runtime-subject.yaml back to activation_status: eval_ready.
2. Keep the new auto-mode method-only fixture as a regression guard.
3. Re-run the accounting eval-ready gate to confirm the fixture pack remains healthy.
4. Defer activation until the failing subject refinement rule is narrowed.