Is your voice agentactually readyto talk to customers?
Paste your prompt, brand context, and email. Get a 0-10 score, 24 items scored, and 3 priority fixes ranked — sent to your inbox in under a minute.
What we grade
line by line.
Each layer scores independently so you know exactly where the prompt is thin. Runtime execution (§SIL, §REP, §ERR) is usually the one that tanks your score — most prompts are all persona, no operational script.
Role, persona, caller context — is your agent a person or a menu?
Turn mechanics, silence handling, error recovery, re-ask limits.
FAQ, objection scripts, lead-qualification fields, pricing guardrails.
TTS rules, response length cap, dynamic variables, prohibited patterns.
State tracking, slot-fill confirmation, repeat caller acknowledgement.
Three steps. One minute.
The full system prompt for your voice agent — persona, rules, tasks, safety instructions.
A paragraph on the brand, target caller, and tone. Helps the reviewer catch missing domain knowledge.
0–10 score, 24 items scored, 3 priority fixes ranked. Shareable results page + optional PDF.
Get monthly voice-AI playbooks.
One email a month. Prompt patterns from production voice agents, common failure modes, teardowns. Unsubscribe anytime.
Everything you might ask.
24 rubric items across 7 layers: agent definition, runtime execution, domain knowledge, output format, memory & state, safety guardrails, and evaluation hooks. Each item gets a 0-10 score with a reason.
Around 30-60 seconds. The scoring engine walks the prompt against every rubric item independently. You'll see live 'scoring…' status while it runs.
No, it's free. We do require an email so we can send you the report and the shareable URL. You can unsubscribe from any follow-up emails in one click; the report is yours to keep.
The prompt is stored so you can share a results URL later, but the results page never publishes your prompt text. If you want the record removed, use the takedown flow linked from the results footer.
Anything that's a system prompt for a voice-first LLM agent — customer support, appointment scheduling, lead qualification, restaurant orders, healthcare intake, real estate. It's not tuned for text-only chatbots, though many of the runtime rules translate.
Codefree Systems & Technologies — an AI-native studio that ships voice AI systems (see the AI Voice & Call Center package). We built this tool because we kept doing manual reviews on client prompts and wanted a repeatable rubric.