oh-my-opencode 5.1.57 → 5.1.59

package/packages/omo-codex/plugin/skills/browser/runtime/omowright/manifest.json

@@ -1,9 +1,9 @@
 {
   "version": "0.0.1",
-  "sourceDigest": "c9eaae75e948cc52d5b60bb654a5565869d68627046453d87d6f695a5b618045",
+  "sourceDigest": "27243299c68de06d4321588d998bd106e0554086d268dc448aa70571db719c1d",
   "files": {
-    "index.js": "8b14c1775f397e36967a0d094c67e2a31f8ae1e92add172925548723ca339046",
+    "index.js": "3a2e4f2575bdfc15448f13768cd8113f8493bf04611c9fefb09b13430d6520de",
     "page-bundle.js": "bfd51e8383cf47200d57387bbbed258bce5a3a3d00c38fc108034d2888e23f0c"
   },
-  "stagedAtUtc": "2026-11-01T20:36:54.907Z"
+  "stagedAtUtc": "2026-11-03T09:03:26.434Z"
 }

package/packages/omo-codex/plugin/skills/eval-integrity/SKILL.md

@@ -0,0 +1,107 @@
+---
+name: eval-integrity
+description: "Assess integrity of an eval, benchmark, metric, A/B test, holdout, LLM-judge, or user-study claim when the question is whether measurement, comparison, or evidence is independent and trustworthy."
+---
+
+# Evaluation Integrity
+
+Provide a read-only audit of whether an evaluation can support its stated claim. Do not alter code,
+data, prompts, labels, or study operations. Begin with exactly one verdict line in the form
+`**Verdict: PASS**`, `**Verdict: FAIL**`, or `**Verdict: INCONCLUSIVE**`. Emit no other verdict
+line and do not restate the verdict later in the report. A favorable score is not itself evidence
+of integrity.
+
+## Route correctly
+
+- **NOT for code review:** route implementation, correctness, security, or maintainability review to
+  `review-work`.
+- **NOT for data processing:** route cleaning, transforming, joining, modeling, or analyzing data to
+  `data-scientist`.
+- Use this skill for the independence and validity of the measurement setup, not for producing a
+  result from it.
+
+## Audit method
+
+### 1. State the claim and map the evaluation
+
+Write the claim in testable form: what comparison, metric, threshold, or causal statement is being
+made; for which population; and under which decision rule. Map every role before judging it:
+
+| Element | Record |
+| --- | --- |
+| Claim | The conclusion and decision it is meant to justify. |
+| System under test | Exact model, version, configuration, prompt, policy, or treatment. |
+| Dataset and split | Sources, collection dates, inclusion rules, train/dev/test or holdout assignment. |
+| Designer | Who selected the task, data, prompts, treatments, metric, and stopping rule. |
+| Evaluator or judge | Who or what assigns quality judgments, including model and prompt versions. |
+| Scorer | The rule or system that turns outputs or judgments into the reported metric. |
+| Subjects | Humans, records, tasks, or units exposed to treatment or observation. |
+
+Trace shared ownership, shared data, shared prompts, shared labels, and shared selection decisions
+across this map. Several models do **not** establish independence merely because there are several
+of them; identify distinct training, tuning, prompt design, labels, data, and decision authority.
+
+### 2. Check the ground truth
+
+Identify the independent ground truth that could confirm the claim without reusing the system's own
+preferences, outputs, labels, or optimization target. Test whether it is:
+
+- external to the system under test and its designer;
+- collected or adjudicated without seeing the treatment identity when blinding is feasible;
+- appropriate to the claim rather than a proxy selected because it favors one arm; and
+- fixed before comparison, or accompanied by a disclosed multiplicity and selection procedure.
+
+If no independent ground truth exists, do not substitute agreement among related systems. Mark the
+claim **INCONCLUSIVE** unless the stated claim is explicitly limited to that related system's score.
+
+### 3. Probe independence failures
+
+Check only risks relevant to the mapped design:
+
+- **Contamination:** benchmark items, answers, or near duplicates reached training, prompt design,
+  retrieval, tuning, or post-hoc selection.
+- **Invalid controls:** arms differ beyond the intended treatment, allocation is biased, exposure is
+  unequal, or the control cannot answer the causal comparison.
+- **Circular judging:** the evaluated system, its designer, or a derived rubric defines its own
+  success without external confirmation.
+- **Scorer-label coupling:** labels, thresholds, reward signals, or rubric choices are reused by the
+  scorer in a way that makes agreement self-fulfilling.
+- **Evaluator-designer coupling:** the evaluator, judge prompt, or adjudication process is selected,
+  tuned, or influenced by the comparison designer without an independent safeguard.
+- **Split or annotator dependence:** overlapping sources, duplicates, temporal leakage, shared
+  annotators, or correlated annotation policies invalidate assumed independence.
+- **Hypothesis priming:** task framing, examples, pre-registration gaps, or interim results steer
+  evaluators or analysts toward the expected outcome.
+- **Observer and demand effects:** subjects or observers infer the desired treatment or outcome, and
+  behavior or ratings change because of that knowledge.
+
+For each concern, distinguish documented evidence from an assumption. Missing provenance for data,
+splits, labels, prompts, evaluator selection, or scoring is **INCONCLUSIVE**, not evidence that the
+design was independent.
+
+### 4. Reach the verdict
+
+- **PASS:** evidence supports an independent, valid measurement path for the stated claim, with no
+  material unresolved dependency.
+- **FAIL:** evidence shows a dependency or invalid design that materially invalidates or overstates
+  the stated claim.
+- **INCONCLUSIVE:** provenance or evidence is missing, the ground truth is not independently
+  established, or remaining uncertainty prevents either result.
+
+Do not convert uncertainty into PASS because a metric looks plausible, a result is statistically
+significant, or multiple related evaluators agree.
+
+## Report
+
+Report the verdict first. Then report only findings that fired; do not pad the report with risks
+that were checked and did not fire. For each fired finding include:
+
+1. **Evidence:** the observed artifact, provenance record, or explicit absence of one.
+2. **Impact:** which claim or inference the issue weakens or invalidates.
+3. **Minimal independence fix:** the smallest separation, blind, held-out source, external
+   adjudication, or control change that would address it.
+4. **Missing evidence:** what must be supplied to resolve the finding, if applicable.
+5. **Residual risk:** what remains uncertain after the minimal fix.
+
+Keep conclusions bounded to the evidence. This audit advises; it does not certify a system,
+recompute results, or approve deployment.

package/packages/omo-codex/plugin/skills/eval-integrity/agents/openai.yaml

@@ -0,0 +1,2 @@
+interface:
+  display_name: "(OmO) eval-integrity"

package/packages/omo-codex/plugin/skills/ultimate-browsing/engine/executor.py

@@ -21,6 +21,7 @@ from __future__ import annotations
 import json
 import os
 import shutil
+import signal
 import subprocess
 import tempfile
 import time
@@ -32,6 +33,10 @@ from .result_schema import Attempt
 
 
 TEMPLATES_DIR = os.path.join(os.path.dirname(__file__), "templates")
+# Chrome removes its per-launch state (the macOS code-sign clone, the temp
+# profile) only on a clean quit, so a timed-out template's process group gets
+# SIGTERM and this long to let `ctx.close()` run before SIGKILL (omo#12067).
+TERM_GRACE_S = 5
 
 
 def _node_available() -> bool:
@@ -70,19 +75,65 @@ def _run_node_template(template: str, args: dict, timeout: int = 90) -> tuple[in
     if not os.path.isfile(path):
         return 127, "", f"template not found: {path}"
     try:
-        proc = subprocess.run(
+        proc = subprocess.Popen(
             ["node", path],
-            input=json.dumps(args),
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
             cwd=TEMPLATES_DIR,
-            capture_output=True,
             text=True,
-            timeout=timeout,
+            start_new_session=True,
         )
-        return proc.returncode, proc.stdout, proc.stderr
+    except Exception as e:
+        return 1, "", f"{type(e).__name__}:{e}"
+    try:
+        stdout, stderr = proc.communicate(input=json.dumps(args), timeout=timeout)
+        return proc.returncode, stdout, stderr
     except subprocess.TimeoutExpired:
+        _terminate_template_group(proc)
         return 124, "", f"timeout after {timeout}s"
     except Exception as e:
+        _terminate_template_group(proc)
         return 1, "", f"{type(e).__name__}:{e}"
+    except BaseException:
+        # Cancellation (KeyboardInterrupt, SystemExit): start_new_session keeps a
+        # terminal Ctrl-C from reaching the template's group, so quit it here.
+        _terminate_template_group(proc)
+        raise
+
+
+def _signal_group(pgid: int, sig: int) -> bool:
+    try:
+        os.killpg(pgid, sig)
+    except ProcessLookupError:
+        return False
+    return True
+
+
+def _terminate_template_group(proc: subprocess.Popen[str]) -> None:
+    """SIGTERM the template's whole group (node and the Chrome it launched), wait up to TERM_GRACE_S, then SIGKILL."""
+    if os.name == "nt":
+        proc.terminate()
+        try:
+            proc.wait(timeout=TERM_GRACE_S)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+    else:
+        pgid = proc.pid
+        if _signal_group(pgid, signal.SIGTERM):
+            deadline = time.monotonic() + TERM_GRACE_S
+            while time.monotonic() < deadline:
+                proc.poll()
+                if not _signal_group(pgid, 0):
+                    break
+                time.sleep(0.05)
+            else:
+                _signal_group(pgid, signal.SIGKILL)
+    try:
+        proc.communicate(timeout=TERM_GRACE_S)
+    except subprocess.TimeoutExpired:
+        proc.kill()
+        proc.communicate()
 
 
 class _FakeResp:

package/packages/omo-codex/plugin/skills/ulw-research/scripts/fidelity-values.mjs

@@ -82,13 +82,16 @@ export function exportRows(ledger) {
 
 const walk = (node) => node.children.filter((child) => child.tag !== "#text").flatMap((child) => [child, ...walk(child)])
 
-/** SVG text nodes of a chart: data labels, or ticks with their declared grid. */
+/** SVG text nodes of a chart: data labels, or ticks with their declared grid; foreignObject HTML counts as unticked labels. */
 function svgLabels(raw) {
-	return walk(parseHtml(raw)).filter((node) => node.tag === "text").map((node) => {
+	const nodes = walk(parseHtml(raw))
+	const texts = nodes.filter((node) => node.tag === "text").map((node) => {
 		let tick = null
 		for (let cur = node.parent; cur && !tick; cur = cur.parent) if (cur.tag === "g" && /(?:^|\s)ticks(?:\s|$)/.test(cur.attrs.class ?? "")) tick = cur.attrs
 		return { text: proseText(node), tick }
 	})
+	const foreign = nodes.filter((node) => node.tag === "foreignobject").map((node) => ({ text: proseText(node), tick: null }))
+	return [...texts, ...foreign]
 }
 
 function onGrid(value, tick) {

package/packages/omo-codex/plugin/skills/ulw-research/scripts/html-foreign-object.mjs

@@ -0,0 +1,61 @@
+// html-foreign-object.mjs - the <foreignObject> carve-out for the html-lite tokenizer.
+//
+// An <svg> element is raw drawing markup, except <foreignObject>, which by definition holds
+// HTML. The tokenizer keeps the svg's inner source in `raw` and additionally parses each
+// foreignObject subtree into real children of the svg node, so the fidelity checks read chart
+// HTML exactly like prose while the rest of the svg stays raw. Plain Node ESM, no dependencies.
+
+import { makeNode, readTag } from "./html-lite.mjs"
+
+/** Newlines in text before offset `to`. */
+function newlinesBefore(text, to) {
+	let count = 0
+	for (let at = text.indexOf("\n"); at !== -1 && at < to; at = text.indexOf("\n", at + 1)) count++
+	return count
+}
+
+/** Shift a parsed subtree's line numbers by `delta` (fragment lines are relative to the fragment). */
+function shiftLines(node, delta) {
+	node.line += delta
+	for (const child of node.children) shiftLines(child, delta)
+}
+
+const isLetter = (ch) => (ch >= "a" && ch <= "z") || (ch >= "A" && ch <= "Z")
+
+/**
+ * Parse an svg node's <foreignObject> subtrees into real children; the rest of the svg stays raw.
+ * @param {object} node an <svg> LiteNode whose `raw` holds its inner source
+ * @param {number} rawStartLine document line of the first raw character
+ * @param {(source: string) => object} parseHtml the tokenizer, passed in so this module never imports it back
+ */
+export function attachForeignObjects(node, rawStartLine, parseHtml) {
+	const raw = node.raw ?? ""
+	for (let lt = raw.indexOf("<"); lt !== -1; lt = raw.indexOf("<", lt + 1)) {
+		if (raw[lt + 1] === "/" || !isLetter(raw[lt + 1] ?? "")) continue
+		const open = readTag(raw, lt + 1)
+		if (open.name !== "foreignobject" || open.selfClosing) continue
+		let depth = 0
+		let innerEnd = raw.length
+		let resume = raw.length
+		for (let at = raw.indexOf("<", open.end); at !== -1; at = raw.indexOf("<", at + 1)) {
+			const closing = raw[at + 1] === "/"
+			if (!closing && !isLetter(raw[at + 1] ?? "")) continue
+			if (readTag(raw, at + (closing ? 2 : 1)).name !== "foreignobject") continue
+			if (closing) {
+				if (depth === 0) { innerEnd = at; const gt = raw.indexOf(">", at); resume = gt === -1 ? raw.length : gt + 1; break }
+				depth--
+			} else depth++
+		}
+		// Tags are lowercased by readTag, so the attached node is "foreignobject" like any parsed tag.
+		const fo = makeNode("foreignobject", open.attrs, node, rawStartLine + newlinesBefore(raw, lt))
+		const frag = parseHtml(raw.slice(open.end, innerEnd))
+		const body = frag.children.find((child) => child.tag === "body") ?? frag
+		for (const child of body.children) {
+			child.parent = fo
+			shiftLines(child, rawStartLine + newlinesBefore(raw, open.end) - 1)
+			fo.children.push(child)
+		}
+		node.children.push(fo)
+		lt = resume - 1
+	}
+}

package/packages/omo-codex/plugin/skills/ulw-research/scripts/html-lite.mjs

@@ -3,10 +3,13 @@
 // Single forward scan with indexOf (no whole-document regex, no backtracking). Produces a
 // LiteNode tree: { tag, attrs, children, parent, line, text, raw }. Text nodes carry
 // tag "#text"; <style>/<script>/<svg> keep their inner source in `raw` and have no parsed
-// children. The "#document" root also carries `ids` (id -> occurrence count) so
-// selectorPath can tell whether an id is unique. Plain Node ESM, no dependencies.
+// children, except an <svg> node also gains its <foreignObject> subtrees as parsed children
+// (see html-foreign-object.mjs; the rest of the svg stays raw). The "#document"
+// root also carries `ids` (id -> occurrence count) so selectorPath can tell whether an id
+// is unique. Plain Node ESM, no dependencies.
 
 import { decodeEntities } from "./entities.mjs"
+import { attachForeignObjects } from "./html-foreign-object.mjs"
 
 export { decodeEntities }
 
@@ -25,7 +28,7 @@ const SAFE_ID = /^[A-Za-z_][\w-]*$/
 export const PROSE_TAGS = new Set(["p", "li", ...HEADINGS, "td", "th", "figcaption", "blockquote", "dd", "dt"])
 export const PROSE_BREAKS = new Set([...P_CLOSERS, ...PROSE_TAGS, "br", "tr"])
 
-function makeNode(tag, attrs, parent, line) {
+export function makeNode(tag, attrs, parent, line) {
 	return { tag, attrs, children: [], parent, line, text: "", raw: null }
 }
 
@@ -181,6 +184,7 @@ export function parseHtml(source) {
 				if (!tag.selfClosing) {
 					const [innerEnd, resume] = findRawClose(src, tag.end, tag.name)
 					node.raw = src.slice(tag.end, innerEnd)
+					if (tag.name === "svg") attachForeignObjects(node, lineAt(starts, tag.end), parseHtml)
 					at = textStart = resume
 				}
 			} else if (!VOID_TAGS.has(tag.name)) {

package/packages/omo-codex/plugin/test/sync-skills-test-support.mjs

@@ -11,6 +11,7 @@ export const expectedSkills = [
 	"comment-checker",
 	"data-scientist",
 	"debugging",
+	"eval-integrity",
 	"frontend",
 	"git-master",
 	"init-deep",

package/packages/omo-codex/scripts/install-dist/install-local.mjs

@@ -1,5 +1,5 @@
 #!/usr/bin/env node
-// omo-codex-install:6cf0b2561cf0c247efdca2934fda04149a66e5051fa585f8eda86922b2c111f2:be8844d5b45ec5d54d7f824d3908bcb0d084368c896fb5cf389ce1699543d2aa
+// omo-codex-install:00fcd516729a06a5b4c6c011662edd0382f55eee0b54a318ea2aad7da009a785:dc6523cbcc0b1a234dee5e4e808122583c5a5341eb73f5ea6bdf421e4d71585c
 var __esm = (fn, res, err) => () => {
   if (fn)
     try {
@@ -10205,7 +10205,7 @@ var package_default;
 var init_package = __esm(() => {
   package_default = {
     name: "@oh-my-opencode/omo-codex",
-    version: "5.1.57",
+    version: "5.1.59",
     type: "module",
     private: true,
     description: "Codex harness adapter for oh-my-openagent. Vendored Codex plugin namespace (omo) + TypeScript installer + telemetry.",
@@ -19693,7 +19693,11 @@ var OmoCategoryConfigObjectSchema = object({
   disable: boolean2().optional(),
   warn_unavailable: boolean2().optional()
 }).strict();
-var OmoCategoryConfigSchema = preprocess((value) => isRecord4(value) ? normalizeLegacyModelFields(value) : value, OmoCategoryConfigObjectSchema);
+function withoutRetiredCategoryKeys(entry) {
+  const { provider_lock: _retired, ...rest } = entry;
+  return rest;
+}
+var OmoCategoryConfigSchema = preprocess((value) => isRecord4(value) ? withoutRetiredCategoryKeys(normalizeLegacyModelFields(value)) : value, OmoCategoryConfigObjectSchema);
 var OmoCategoriesConfigSchema = record(string2(), OmoCategoryConfigSchema);
 
 // packages/omo-config-core/src/schema/computer.ts