fix(ci): fail zsh benchmark on red thresholds

- Add a zsh-bench threshold check that matches the dashboard red status rules
- Stop using github-action-benchmark relative alerts as the CI failure condition
- Apply the same red-threshold check to main and PR benchmark workflows

Co-authored-by: Codex GPT-5.5 <noreply@openai.com>

Author: ve62ca

.github/scripts/check-zsh-bench-thresholds.sh

@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+# Fail CI only when a zsh-bench metric reaches the dashboard red state.
+set -euo pipefail
+
+input="${1:-benchmark-result.json}"
+
+if [[ ! -r "$input" ]]; then
+  echo "::error::Benchmark result file not found: $input"
+  exit 1
+fi
+
+failures="$(
+  jq -r '
+    def targets:
+      {
+        "first prompt lag": 50,
+        "first command lag": 150,
+        "command lag": 10,
+        "input lag": 20
+      };
+
+    .[]
+    | select((targets[.name] // null) != null)
+    | . as $metric
+    | (targets[$metric.name]) as $target
+    | ($target * 2) as $red_threshold
+    | select(($metric.value | tonumber) > $red_threshold)
+    | [
+        $metric.name,
+        (($metric.value | tostring) + " " + ($metric.unit // "ms")),
+        (($target | tostring) + " ms"),
+        ("> " + ($red_threshold | tostring) + " ms")
+      ]
+    | @tsv
+  ' "$input"
+)"
+
+if [[ -z "$failures" ]]; then
+  echo "All thresholded zsh-bench metrics are below dashboard red status."
+  exit 0
+fi
+
+echo "Red zsh-bench metric(s) detected. Dashboard red status fails CI."
+echo "metric | value | target | red threshold"
+echo "--- | --- | --- | ---"
+
+while IFS=$'\t' read -r metric value target red_threshold; do
+  echo "${metric} | ${value} | ${target} | ${red_threshold}"
+  echo "::error title=Zsh bench red metric::${metric} is ${value}; red threshold is ${red_threshold}."
+done <<< "$failures"
+
+exit 1

.github/workflows/benchmark-pr.yml

@@ -37,12 +37,14 @@ jobs:
           github-token: ${{ secrets.GITHUB_TOKEN }}
           auto-push: false
           save-data-file: false
-          alert-threshold: "150%"
-          comment-on-alert: true
           comment-always: true
-          fail-on-alert: true
+          fail-on-alert: false
           summary-always: true
 
+      - name: Fail on red zsh-bench metrics
+        shell: bash
+        run: bash .github/scripts/check-zsh-bench-thresholds.sh benchmark-result.json
+
       - name: Clean benchmark temporary files
         if: always()
         shell: bash

.github/workflows/benchmark.yml

@@ -58,12 +58,13 @@ jobs:
           benchmark-data-dir-path: .
           github-token: ${{ secrets.GITHUB_TOKEN }}
           auto-push: true
-          alert-threshold: "150%"
-          comment-on-alert: true
           comment-always: true
-          fail-on-alert: true
+          fail-on-alert: false
           summary-always: true
-          alert-comment-cc-users: "@lemtoc"
+
+      - name: Fail on red zsh-bench metrics
+        shell: bash
+        run: bash .github/scripts/check-zsh-bench-thresholds.sh benchmark-result.json
 
       - name: Clean benchmark temporary files
         if: always()