Harlo Agent Eval Kit Pro — golden cases, rubrics & CI gate for AI agents
Catch agent regressions before your users do. Harlo Agent Eval Kit Pro ships ready-to-run evaluation packs for teams shipping AI agents and chatbots.
What’s included
• 5 agent-type packs (108 golden cases): support-bot, lead-qual, rag-docs, tool-calling-ops, personal-assistant — JSONL + CSV
• 12 rubrics on a 1–5 scale with pass / partial / fail examples
• Strict-JSON judge prompt with bias mitigations
• One promptfoo config per pack (offline echo provider — no API keys to try the wiring)
• GitHub Actions regression gate: pack validation + config validation + judge calibration (green by default); per-pack echo eval is informational until you plug in your agent
• Judge calibration script + sample human labels
• Pro Guide (PDF + Markdown) and for-agents.md
• No-code CSV scoring sheet
Evals are a quality process, not a guarantee of agent behavior. All companies, people, and figures in the packs are fictional. “for Muse” is a descriptive label only; Harlo is not affiliated with, endorsed by, or sponsored by Meta.
Brand: Harlo
Support: digitalpromohub.support@gmail.com
Free Lite (support-bot pack, MIT) on GitHub: https://github.com/miketusa-ctrl/evalkit-lite
Also available: Team ($99, client-work license + pack-authoring guide) on this storefront.
Pairs well with Harlo Support Kit, Lead Qual Kit, and Security Kit.
How delivery works
- Buy on Payhip
- Download harlo-evalkit-pro.zip
- Follow QUICKSTART.md / docs/PRO-GUIDE.md
How do refunds work?
Digital download — sales are final; if you can’t access the zip, contact support through the store and we’ll help or make it right.