Tribunal · Overview

Status — Tribunal vs Codex Tribunal

Two distinct things share this name. Disambiguate before reading further:

• Codex Tribunal (mentioned in Field Proof v0.1) is a real, external, independent audit process that audited the field-release prototype.
• The in-product Tribunal described on this page is the internal rulebook engine that runs inside DefendableOS — the referee.

Earlier docs described the in-product Tribunal as a “two-judge ensemble” with a 6-dimension weighted scoring model. That framing is superseded as of the 2026-05-27 doctrine lock. The in-product Tribunal is a deterministic rulebook engine, not a judge model.

What the in-product Tribunal is

The in-product Tribunal is the rulebook engine inside DefendableOS — the referee that applies a declared Flight Sheet to an agent submission and throws flags when a rule is violated. Every check passes or raises a flag. There is no opinion grade. There is no “seems good.” Score = % of declared rules satisfied.

“The referee is a rulebook, not a judge.” — Mr. Defendable, 2026-05-27

See Rulebook Engine for the full operational detail. The pages in this section document the legacy framing surfaces (scoring model, validator chain, verdict JSON) reframed to the new doctrine.

What the referee does, briefly

1. Run the rulebook. Apply each declared rule from the Flight Sheet.
2. Throw flags. Each check passes or raises a flag with a tier (low/mid/high) and severity (critical → propolis, noncritical → jelly).
3. Emit a deterministic verdict. Score = % of declared rules satisfied. Severity = honey (no flags + approved) · jelly (mid/low only) · propolis (any high-tier flag). Risk breakdown = count of flag tiers.

No LLM is called on the receipt path. A judge-model slot is reserved for advisory hallucination/readiness signals later — never on the chain.

The pages in this section

• Scoring Model — score = % of declared rules satisfied; the 6-dim weighted opinion model is superseded.
• Honey · Jelly · Propolis — the 3 severities the referee emits.
• Risk Temperature — the low/mid/high tier system used to rank flags.
• Evidence Strength — evidence is a set of declared rules (non-empty, cited, labeled), not a 0-1 dial.
• Validator Chain — the deterministic sequence the referee runs.
• Verdict JSON — the live verdict shape.
• Assignment Success — what “success” means under the rulebook (all required rules satisfied).

---

🐝 Rulebook · flags · receipts. The referee is a rulebook. To the shed.