Version packages

Author: CharGrnmn

.changeset/forty-dingos-chew.md

@@ -1,5 +1,16 @@
 # @razroo/iso-eval
 
+## 0.3.0
+
+### Minor Changes
+
+- 5f4178f: Add a built-in Codex runner to `iso-eval` so suites can execute against a
+  real coding harness instead of only the fake runner.
+
+  Expand `iso-trace` with Codex and OpenCode transcript support, session
+  discovery across those harnesses, `model-score` reporting, and
+  CI-friendly routing/schema regression gates for OpenCode tool usage.
+
 ## 0.2.0
 
 ### Minor Changes
@@ -19,11 +30,11 @@
   - type: agentmd_adherence
     promptFile: ../agent.md
     fixtures: ../fixtures.yml
-    ruleId: H3               # optional
+    ruleId: H3 # optional
     minPassRate: 0.9
-    via: claude-code         # optional (api | claude-code | fake)
-    model: claude-haiku-4-5  # optional
-    timeoutMs: 180000        # optional
+    via: claude-code # optional (api | claude-code | fake)
+    model: claude-haiku-4-5 # optional
+    timeoutMs: 180000 # optional
   ```
 
   Adds `@razroo/agentmd` as a runtime dependency so installing

package-lock.json

@@ -1,6 +1,6 @@
 {
   "name": "@razroo/iso-eval",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "description": "Behavioral eval runner for AI coding agents — snapshot a workspace, hand it to a runner with a task prompt, score the resulting filesystem/git state.",
   "license": "MIT",
   "type": "module",

packages/iso-eval/CHANGELOG.md

@@ -1,5 +1,16 @@
 # @razroo/iso-trace
 
+## 0.3.0
+
+### Minor Changes
+
+- 5f4178f: Add a built-in Codex runner to `iso-eval` so suites can execute against a
+  real coding harness instead of only the fake runner.
+
+  Expand `iso-trace` with Codex and OpenCode transcript support, session
+  discovery across those harnesses, `model-score` reporting, and
+  CI-friendly routing/schema regression gates for OpenCode tool usage.
+
 ## 0.2.0
 
 ### Minor Changes
@@ -29,9 +40,9 @@
 ### Minor Changes
 
 - Initial release. Parse Claude Code JSONL session transcripts into a
-  harness-agnostic event model so you can ask *which rules ever
-  actually fired?*, *which tools does my agent reach for most?*, and
-  *which captured sessions would make good regression fixtures?* Ships
+  harness-agnostic event model so you can ask _which rules ever
+  actually fired?_, _which tools does my agent reach for most?_, and
+  _which captured sessions would make good regression fixtures?_ Ships
   `sources`, `list`, `show`, `stats`, and `export` commands. Zero
   upload — everything is local reads and user-controlled output. Codex
   and OpenCode parsers are additive and planned for later minor

packages/iso-eval/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@razroo/iso-trace",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "description": "Parse, normalize, and query AI coding agent transcripts — Claude Code JSONL sessions first, Codex and OpenCode additive. Local-only; never uploads.",
   "license": "MIT",
   "type": "module",