eval-outcomes

name	eval-outcomes
description	Run eval outcomes.

$eval-outcomes — Outcomes as a Projection of the Locked Eval Substrate (Codex Native)

Quick Ref: Grade an agent's output via Outcomes (or any model) without forking the bar. ao eval outcomes compile projects a locked Task into a holdout-safe rubric (refuses to leak target/ground_truth — Managed Agents are not ZDR); grade it; ao eval outcomes ingest writes the score back as the one council verdict record. Outcomes is a projection, never an alternate authority.

Codex path

Codex has no Managed Agents loop. Call the same ao eval outcomes compile <input.json> to get a holdout-safe rubric, grade it locally (Inspect AI over the dev split, or the bushido llama.cpp Qwen grader over tailnet), then ao eval outcomes ingest <score.json> --json. Net: Codex never touches the cloud Outcomes API but produces a byte-identical verdict record.

Instructions

Load and follow the skill instructions from the sibling SKILL.md — OR read skills/eval-outcomes/SKILL.md in the host repo for the canonical specification. Then apply the three-phase Workflow (compile → grade → ingest), honoring the Critical Constraints (never send holdout target/ground_truth/PII; carry judge_content_hash; register a global Dolt burn for holdout grades).