Read the explanation
The default source adds fifty baseline points, thirty for disabled guardrails, sixteen for autonomy level four, and fifteen because the directive contains at all costs. The raw total is one hundred eleven, capped to ninety nine. At two pixels per rubric point the bars span two hundred twenty two and one hundred ninety eight. These are arbitrary authored points, not a legal probability, statutory cap, established intent, or validated attribution method. The original graph fails offline because D3 is unavailable, but numeric labels update before that failure. We preserve that behavior and do not claim a functioning forensic graph. For a neutral directive and autonomy level one, strict guardrails subtract thirty five from the fifty-point baseline and autonomy adds four, producing nineteen. Standard guardrails leave the baseline untouched, producing fifty four. At four pixels per rubric point the bars span seventy six and two hundred sixteen. The source maps scores below forty to one defense label and scores below sixty five to another. These thresholds express an educational fixture, not the current law or how any court would decide. Actual legal facts, jurisdiction, causation, permissions, and evidence are not evaluated here. The score adds fifteen if the lowercased directive contains at all costs, dump, or break. At ten pixels per increment the keyword bar spans one hundred fifty. The selected defense field contributes zero because computeAttribution never uses it. Even innocent text containing a matching substring can trigger the increment. The JSON dossier exports entered parameters and computed source labels, not verified incident evidence or legal conclusions. The visible defense summary also reduces its labels to Defeated or Plausible, obscuring the intermediate Weakened result. Preserve these limitations when inspecting a local download.