feat(fase3): additive meter= seam on run_project (SC3 detach hook)

tests/test_vertical_slice_e2e.py

@@ -199,6 +199,39 @@ async def test_h_tiny_budget_halts_via_debate_middleware(
         )
 
 
+async def test_i_injected_meter_is_used(docs_dir, make_client_factory) -> None:
+    """Step 1 (SC3 detach hook): a TokenMeter passed as ``meter=`` IS the meter the run uses —
+    its pre-charged tokens flow through to ``provenance.token_usage`` (run.py:179). Detaches
+    cleanly: reverting the run.py:151 injection makes the injected instance be ignored (a fresh
+    internal meter is built instead), the pre-charge vanishes, and the equality below fails.
+    Pattern: test_a (run_project call shape)."""
+    from portfolio_optimiser.budget import Budget, TokenMeter
+
+    factory = make_client_factory(_VALID)
+    baseline = await run_project(
+        "FV42-GSV-E1",
+        "local",
+        docs_dir=docs_dir,
+        verdict_input=_VI,
+        client_factory=factory,
+    )
+    pre_charged = 5000
+    injected = TokenMeter(Budget(max_tokens=100_000, max_rounds=max(3 * 4, 4)))
+    injected.charge(pre_charged)
+    result = await run_project(
+        "FV42-GSV-E1",
+        "local",
+        docs_dir=docs_dir,
+        verdict_input=_VI,
+        client_factory=factory,
+        meter=injected,
+    )
+    assert baseline.provenance.token_usage > 0  # the run charges a deterministic delta
+    # The injected instance flows through: end tokens == pre-charge + the same run delta.
+    assert result.provenance.token_usage == pre_charged + baseline.provenance.token_usage
+    assert result.provenance.token_usage > baseline.provenance.token_usage
+
+
 async def test_f_malformed_contract_raises_before_any_chat(docs_dir, make_client_factory) -> None:
     calls = {"n": 0}
     base = make_client_factory(_VALID)