predicate: Instructions file for Codex

AGENTS.md — Predicate

This repository is Predicate, the upstream. skills/, rules.md, and ambient.md live at the root because other projects consume them as an installed plugin — see README.md and docs/getting-started.md.

Goal

Predicate is a self-hosting, harness-agnostic framework that keeps AI coding agents anchored to the true goal across long-horizon work. It has two synergistic halves:

• Correction — externalize correctness to the strongest evaluator (the Verification Dual: verify, then trust — no condition closed by an agent's say-so). Authority: rules.md.
• Prevention — externalize the goal, requirements, unknowns, and available tools as a durable, selectively-projected conditioning layer (the always-on conditioning/ system-prompt law, the /orient workflow, the nested AGENTS.md hierarchy), so a walk stays focused before drift compounds.

Correction without prevention corrects toward the wrong goal; prevention without correction drifts anyway over a long horizon. Together they bound drift.

Directions

A direction is a coherent, non-terminal motivation the project moves toward and never completes. Directions are deliberately few; minting a new one is reserved to a human ruling, while serving an existing one is free. They are distinct from the goal hierarchy above: that hierarchy is containment (ecosystem ⊃ project ⊃ component) and governs where a defeater escalates, while directions are concern and form a lattice — a single piece of work can serve more than one, and cross-cutting work is usually where the leverage is. Work does not converge on a direction; it converges on a terminal claim, a checkable statement that, once corroborated, means that line of work is finished. A terminal claim carries the same backing discipline as any other claim under the Verification Dual — "is this done" is answered with evidence, never a bare assertion — and the fraction of a direction's terminal claims that have been discharged is the project's convergence signal.