compliance-gate-spot-check

Compliance Gate Spot-Check

Purpose: 77+ structural-RTL gates may report PASS but include gameable patterns where:

• Gate's regex is too narrow (Wave-30/33 have history of this — slave_tx_no_device_break_check originally missed tx_oe_low because \b boundary)
• Project carries a waiver that satisfies the gate but the underlying defect is real
• The gate's PASS condition is satisfied trivially (e.g. "L9 has ≥3 typed fields" satisfied by stub fields)

Prioritize the sampling budget (optional — falls back to uniform sampling)

Before sampling, ask the gate-reliability register which gates have a history of being gamed / false-passed, so the limited spot-check budget lands on the historically-dangerous gates first instead of uniformly across all 77. The register is programs/gate_reliability_register.py — a self-calibrating per-gate EMA ledger of pass-rate and false-PASS-rate. It is an optional prioritizer: when the ledger is empty/absent it ranks nothing, and you fall back to the uniform "5 random gates" behavior in step 1 below.

The ledger path is positional (the program calls it LEDGER). Use a stable location such as <project>/reports/gate_reliability_register.json (or a shared cross-project ledger). Steps:

1a. Ensure the ledger exists, then rank. A freshly touched (empty) ledger ranks to [] and you proceed uniformly — so this never blocks you:

```bash
# create on first use so `rank` never errors on a missing file
touch <project>/reports/gate_reliability_register.json

# priority-ordered gate list (highest false-PASS history first)
python3 programs/gate_reliability_register.py rank \
    <project>/reports/gate_reliability_register.json --top 10
```

The output is a JSON array of `{"gate": ..., "spotcheck_priority": ...}`
sorted highest-priority first. **Spend your sampling budget on the
top-ranked gates** (they are the historically-gamed ones) before sampling
the remaining gates at random. If the array is empty, sample uniformly.