[shrsv]

Adds a read-only existence check that avoids copying the stored value.

Simulated for this demo: org risk calibration is risk_threshold_percentile: 80 (lenient), calibrated from this org's own recent score distribution. This diff's risk score lands around 0.55-0.6 -- moderate, but below this org's cutoff.

[livereviewbot]

LiveReview automated review

Scanned 1 changed file (+8/-0). New read-locked accessor, mirrors the existing Get pattern. No findings; moderate risk score (~0.55-0.6) reflecting that it's a new public method on a shared type, even though the logic itself is simple.

[livereviewbot]

Auto-approve policy evaluation

Stage	Result
Eligibility	✅ pass
LLM classification (refactor-only, confidence 9)	✅ pass
Org risk calibration	risk_threshold_percentile: 80, computed against this org's last 10 review scores: [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.65, 0.7, 0.8, 0.9] → P80 cutoff ≈ 0.8
Risk score vs. cutoff	✅ ~0.57 < 0.8

Verdict: AUTO-APPROVED. This org has dialed its auto-approve calibration lenient -- it auto-approves anything below the 80th percentile of its own recent review scores. Compare with threshold-p10-strict-blocks-same-score: identical diff, identical score, opposite outcome, because the org dial is different.

[livereviewbot]

Auto-approve policy verdict: APPROVED. Risk score (~0.57) sits below this org's lenient P80 calibration cutoff.