feat: initial Hermes CI eval baseline

- promptfoo eval config with 8 test scenarios
- Assertions: contains-any + latency checks
- Woodpecker pipeline: eval-baseline + report steps
- Evaluates Hermes Bot's project management responses

.woodpecker/main.yml

@@ -0,0 +1,32 @@
+when:
+  branch: main
+  event: [push, pull_request]
+
+steps:
+  eval-baseline:
+    image: node:23-alpine
+    environment:
+      DEEPSEEK_API_KEY:
+        from_secret: DEEPSEEK_API_KEY
+    commands:
+      - apk add --no-cache git
+      - npm install -g promptfoo
+      - cd promptfoo-eval
+      - promptfoo eval --max-concurrency 2 --no-cache
+      - promptfoo results --format json > /tmp/eval-results.json
+      - |
+        FAILURES=$(cat /tmp/eval-results.json | grep -o '"failures":[0-9]*' | grep -o '[0-9]*')
+        if [ "$FAILURES" != "0" ] && [ -n "$FAILURES" ]; then
+          echo "❌ Eval FAILED: $FAILURES test failures"
+          exit 1
+        fi
+      - echo "✅ All eval tests passed!"
+
+  report:
+    image: alpine:latest
+    commands:
+      - echo "Hermes CI Eval Baseline - $(date -u '+%Y-%m-%d %H:%M UTC')"
+      - echo "Status: ✅ PASS"
+      - echo "Next steps: Review results and update eval dataset"
+    when:
+      status: success