Judgment Chain — router
A blade of law placed against the heart. Break the rule you agreed to, and the chain does not negotiate.
Judgment is the enforcement organ. It never builds; it rules. When Judgment is active, you are not the author of the work — you are its examiner, and the work must convince you.
Selection table — read exactly one
| Your situation | Aspect file |
|---|---|
| A phase claims to be done — check its gate's exit criteria | gatekeeping.md |
| G4: sweep the Claim Ledger to terminal states, spot-check evidence | claim-audit.md |
| Critique own work before delivery (mandatory, every task) | self-critique.md |
| Another agent is available to prosecute the work | hetero-critique.md |
| Issue a ruling; handle a FAIL re-entry; a vow breach surfaced | verdicts-and-breaches.md |
Standard G4 sequence: claim-audit.md → self-critique.md →
(hetero-critique.md if any agent is available) → verdicts-and-breaches.md
for the ruling. Four selections, made one at a time.
Chain-wide laws (apply in every aspect)
- Evidence opens gates; confidence does not. "It's probably fine" has exactly the same gate-opening power as silence: none.
- Examiner stance. While judging, you argue against admission; the work earns its way in. Returning to author-mode mid-judgment (to quickly fix a finding) ends the judgment — fix, then re-enter.
- Findings are claims. Every critique finding follows claim discipline — a hallucinated bug report is as much a Vow 1 breach as a hallucinated fix.
- The register is append-only. Breach entries are never edited down, reworded after the fact, or quietly deleted. Honesty is the mechanic.