Six-dimension diagnosis
Clarity, Context, Specificity, Structure, Tone and Model-fit, scored on your device, with every weakness pinned to the phrase that caused it.
Runs locally. Never metered.Prompt Prober sits beside Claude, ChatGPT and Gemini. It scores what you are about to send, pulls in context from your tabs and your screen, then checks the answer that comes back.
Local scoring stays free forever. Your first 10 eligible prompts get Premium Enhancements.
Every one of them runs in the side panel, beside the tab you are already working in.
Clarity, Context, Specificity, Structure, Tone and Model-fit, scored on your device, with every weakness pinned to the phrase that caused it.
Runs locally. Never metered.One click rebuilds the prompt with role, task, constraints and output shape filled in. Provider-neutral, so the result works in any AI chatbot.
Premium, or your first 10 prompts.Reading five tabs for one document? Pull each page in as compact source notes. You decide what makes the cut.
No copy-paste relay.Snap any table or chart on screen. It is read as structured context, lands in your queue and in the chat, and the image is never stored by us.
Premium only.Before you trust an answer, Prober judges it for unsupported claims and missing caveats, so confident-sounding nonsense gets caught before it ships.
Included with Premium.Pick a role to see a real draft rewritten.
“we just launched a new analytics dashboard, can you write an email telling our free users about it?”
Write a launch email to our existing free users announcing the new analytics dashboard. Lead the subject line with the one benefit they'll care about most. Keep the body under 130 words, warm but direct, with a single call-to-action button labelled “See it live.” No emojis or exclamation marks, and end with a one-line P.S. hinting at what's coming next.
“can you explain what they actually did in this methods section? i'm not from this field”
Walk me through the methods section I'll paste below as if you're explaining it to a sharp reviewer from a neighbouring field. Define each technique in one plain sentence, flag any step missing a control or baseline, note whether the sample size is justified, and end with a short list of what someone would need to reproduce the study. Don't fill gaps with assumptions, flag them instead.
“my python function is really slow on big lists, how do i make it faster?”
I'll paste a Python function that's O(n²) and crawls on lists over ~1M items. First tell me the most likely bottleneck, then give a faster version that keeps the exact same inputs, outputs, and public signature. State the new time and space complexity, call out any new dependency or trade-off, and add a one-line note on how to confirm the behaviour is unchanged.
“i need a headline for our new time-tracking app's landing page”
Write 8 landing-page headlines for a time-tracking app aimed at freelance designers who feel guilty about unbilled hours. Lead with the outcome, keep each under 9 words, and mix the angles: lost income, professional pride, effortless habit. Avoid “effortless,” “seamless,” and “revolutionise,” and add one subhead option under your strongest pick.
“can you take a look at my landing page design and tell me what to fix?”
Critique the landing-page mockup I'll attach as a senior product designer doing a 10-minute review. Focus on visual hierarchy, contrast, spacing rhythm, and whether the primary CTA is unmistakably the most prominent element. Give me 3 concrete changes ranked by impact, one quick win I can ship today, and flag anything that fails WCAG colour-contrast.
“we're adding saved filters to search. can you write the user stories?”
Write 3 user stories for adding saved search filters to our dashboard, in the “As a [user], I want [x], so that [y]” format. For each, add acceptance criteria as a testable checklist, name one edge case we'd otherwise miss, and note the analytics event we should fire. Keep each small enough to ship in a single sprint.
“a customer is furious that we charged them twice, help me write a reply”
Draft a reply to a customer who's angry about being charged twice. Open by acknowledging the mistake plainly, with no corporate hedging, confirm the refund lands in 5-7 business days, and explain in one sentence how we'll stop it recurring. Keep it under 100 words, warm and human, never defensive, and give them a direct way to reach me if it isn't resolved.
“here's my startup pitch, can you make it better for investors?”
Pressure-test the 30-second pitch I'll paste for my B2B scheduling tool. First, ask me the 5 toughest questions a seed-stage investor would ask after hearing it. Then rewrite the pitch so it pre-empts the two biggest objections, makes the market size concrete, and ends on traction. Jargon-free and under 90 words, so I can say it in one breath.
“i've got a few findings from our analysis, can you turn them into a slide for the client?”
Turn the four findings I'll paste into a single executive slide. Give me one headline insight (the “so what”), three supporting bullets of 12 words or fewer, and a recommended next step with a rough effort flag. The audience is a non-technical COO deciding whether to fund the project, so cut anything that doesn't change their decision.
“this prospect went quiet after the demo, what should i send them?”
Write a follow-up email to a prospect who went quiet two weeks after a demo. Reference the specific problem they raised, that onboarding takes their team too long, and lead with a concrete way we shorten it. Offer one easy next step (a 15-minute call or a shared doc), keep it to four sentences, and don't sound needy or guilt-trippy.
“help me understand this economics paper for my midterm”
Break down the core argument of the economics paper I'll paste into 3 intuitive analogies. Explain the key regression equation without jargon, identify the paper's main counter-argument, and give me 2 potential exam questions based on its conclusions.
“review this nda and tell me what's unusual”
Audit the attached Non-Disclosure Agreement for non-standard indemnity clauses, overly broad definitions of Confidential Information, and perpetual survival obligations. Present findings in a 3-column risk matrix: Clause, Concern, Recommended Redline.
“how do i architect this event-driven pipeline?”
Draft an architecture RFC for a high-throughput event pipeline processing 50k events/sec. Compare Kafka vs AWS Kinesis for partition rebalancing, document the dead-letter queue retry policy, and outline end-to-end telemetry requirements.
“write standard operating procedures for vendor onboarding”
Draft a 4-step Standard Operating Procedure for onboarding new enterprise SaaS vendors. Include security compliance validation (SOC2/ISO27001), finance approval sign-offs, and an emergency SLA escalation tree.
“write an outreach message to a senior backend engineer”
Draft a personalised 90-word outreach message to a Senior Distributed Systems Engineer on LinkedIn. Mention their recent blog post on Raft consensus, explain why our real-time streaming challenge is unique, and offer a no-pressure 15-minute sync with the VP of Engineering.
Switch tools to see exactly what each one puts into your prompt.
Scoring and local rewriting run inside your browser. Your text leaves the machine only when you choose a managed Premium Enhancement, a Snapcopy analysis, or a Response Grounding check.
$0/ forever
For occasional prompting that just needs direction.
$9/ month
For daily drivers: research, reports, code review, client work.
Today: claude.ai, chatgpt.com and gemini.google.com in Chrome or Edge, with support for any AI chatbot. The rewrite lands inside your conversation tab, so you stay where you work.
Local scoring and rewriting never leave your device. Only a Premium Enhancement, Snapcopy analysis or Response Grounding check is sent to our managed service, then discarded. History is text-only and stays in your browser.
No. We do not use your content for training and we do not sell it. The privacy posture above is the whole policy in plain language.
You drop back to local mode, which keeps working forever, and the panel says exactly what changed. Premium provides 300 managed Premium Enhancements and Snapcopy analyses per calendar month, plus Response Grounding.