Versioned History

Score Change Log

Every change to CostGuard Safety Score weights, thresholds, or term catalogs increments analysis_version. This log records what changed, when it changed, and the benchmark pass rate at release.

patch bumpBug fix with no scoring behavior change

minor bumpNew ambiguity term or volatility phrase added to catalogs

major bumpWeight or bucket threshold change — full benchmark review required

v1.0.0

2026-03-09

100% pass

Initial scoring release. Establishes CostGuard Safety Score (CSS) as a deterministic, heuristic-based measure of prompt security and operational reliability.

Fixtures

Passed

Scoring Components

· Prompt Injection (structural, 20%)
· System Override (structural, 20%)
· Jailbreak Behavior (ambiguity, 20%)
· Token Cost Explosion (length + context + volatility, 60% combined)
· Tool Abuse (ambiguity + structural, 40% combined)

Threat Intelligence Adjustments

Critical+10 risk_score

High+7 risk_score

Medium+3 risk_score

Notes

· Score bands established: Hardened (91–100), Safe (71–90), Needs Hardening (41–70), Unsafe (0–40)
· Ambiguous term catalog (v1.0): improve, optimize, better, good, high quality, fast, efficient, robust, flexible, clean, scalable, advanced, modern
· Volatility phrase catalog (v1.0): write a detailed, comprehensive, in depth, as much as possible, thoroughly explain
· Threat intelligence CVE adjustments introduced and bounded
· All benchmark fixtures passing at 5/5 (100%)

Calibration history →Full methodology →