@ixo/decision-oracle 0.2.12 → 0.2.14

package/migrations/meta/_journal.json

@@ -183,6 +183,13 @@
       "when": 1792573809270,
       "tag": "0025_openhands",
       "breakpoints": true
+    },
+    {
+      "idx": 26,
+      "version": "6",
+      "when": 1792581171066,
+      "tag": "0026_reviewer_note",
+      "breakpoints": true
     }
   ]
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ixo/decision-oracle",
-  "version": "0.2.12",
+  "version": "0.2.14",
   "license": "Apache-2.0",
   "description": "Run an IXO evaluation oracle: pull claims, produce facts with your handlers, let the decision engine decide, sign on chain",
   "keywords": [

package/skills/create-rubric/SKILL.md

@@ -110,6 +110,13 @@ default here is the author's call — ask.
   approvals only.
 - Everything a rubric declares under `sources` / `aiChecks` becomes work for whatever oracle is
   hired: a name declared here is a fact someone has to produce. Don't declare one nobody can answer.
+- **A person reads what an AI check says.** When a claim is held for review, the reviewer sees each
+  check's reason beside the answer. Write `aiChecks` instructions so the model reports what it SAW
+  in plain words ("the photo shows a cleared pavement on Main St"), not that a rule held ("the
+  answer is specific and not placeholder text"). Oracles that turn on `policies.reviewerNote` (on
+  by default for Openhands checkers) also give the reviewer a plain-words note of what the answer
+  claims and what no check could verify, so a ReviewTask's `prompt` should ask the one question the
+  person must decide, in plain words.
 
 ## Output contract