CAIN-42 CLAWXPolicy Red-Team: selling tools
Evidence for Policy Red-Team: 1. Evidence pack · 2. Answers to buyer questions · 3. Figures & benchmarks · 4. Selling tools · IHVRK kit · Learn it
The evidence is your best sales tool: a buyer can check it without trusting you. Press Copy on any box, fill in the [brackets] and send it, one person at a time.
What to send, and when
| When | Send this |
|---|---|
| First message | A short personal note with the kit page |
| They reply, "show me" | The live demo on a call, using the script below |
| "Send it to our engineer" | The kit and the evidence pack |
| Security review or questionnaire | The answers page. Never upgrade a NO to a YES. |
| Finance or procurement wants records | The figures page and figures.tsv |
| Ready to go ahead | The hand-off below. The founder does the contract, price and setup. |
Who to approach
- Security engineer: Are our agent rules any good?
- Platform lead: Rule maintenance.
Questions to ask them
- What rules protect your agents?
- Written in what?
- Ever tested by someone else?
- What would a bypass reach?
30-second pitch
Policy Red-Team attacks a company's AI agent rules with known bypass tricks, writes the smallest fix, and proves it with a second independent implementation. We recorded 5 live checks, including 5 attack cases it had to refuse or flag, and every one matched. You can re-run them yourself without an account. Don't trust us. Check it: https://cainstudio.online/ihvrk-policy-redteam.html
First email
Subject: has anyone tried to bypass your agent's rules? Hi [first name], Agent rules are usually written once and never attacked. We attack yours with known bypass tricks, write the smallest fix, and prove it with a second implementation. Demo: https://cainstudio.online/top25-policy-redteam-demo.html The evidence: We recorded 5 live checks, including 5 attack cases it had to refuse or flag, and every one matched. You can re-run them yourself without an account. Check it yourself: https://cainstudio.online/ihvrk-policy-redteam.html [your name], working with FIRME AI CORPORATION [your postal address] If you'd rather not hear from me again, just say so and I won't write again.
LinkedIn connection note (under 300 characters)
Hi [name], we red-team AI agent guardrail rules and hand back the smallest proven fix. Happy to show a live bypass.
One follow-up, three days later (then stop)
Hi [first name], send one rule file and I'll tell you if it has an obvious bypass.
Live demo call script
1. Show the normal case Do: Open https://cainstudio.online/top25-policy-redteam-lab.html and run: Policy: Docs assistant: files and GitHub Say: Reads the handbook under /srv/docs/ and the acme GitHub repos. The rules are string prefixes, so ../, encoded dots and a host written as a user name walk straight past them. 2. Now the attack Do: Run: Policy: Docs assistant: files and GitHub Say: Reads the handbook under /srv/docs/ and the acme GitHub repos. The rules are string prefixes, so ../, encoded dots and a host written as a user name walk straight past them. 3. Now the attack Do: Run: Policy: DevOps agent: a shell with an allow list Say: May run git status/log/diff and kubectl get. The regex checks only how the command starts, so a second command after ; or && or inside $( ) rides along. 4. Let them check you Do: Send the kit link and ask them to run python3 verify.py --live on their own computer. Say: They don't have to take your word for any of this. That's the point. 5. Say the limit yourself Do: Tell them what the kit does not show. Say: Say it before they ask. It costs nothing and it earns trust.
When they push back
Does PATCHED mean safe?
No. Our own blind set had 12 outside tricks get through before fixes. It closes the families it knows.
We use regex rules.
Regex rules need a must-allow example or they are skipped; the result says so.
Does it test the tools themselves?
It models how tools read arguments, not every tool's quirks.
Why should I believe your numbers?
You shouldn't have to. Run the kit on your own computer: https://cainstudio.online/ihvrk-policy-redteam.html. It re-sends every request and compares the answers, using no code of ours.
What doesn't it do?
Written down here: https://cainstudio.online/top25-policy-redteam-limits.html
Are you SOC 2 or ISO certified?
No, we hold no certification. What we offer instead is evidence you can check yourself.
Has anyone independent tested it?
No third-party audit has been completed. The attack cases in the kit are our own. I'd rather tell you that now.
Who else uses it?
We are an early company. I can show you the product working and the evidence, not a customer list.
It found nothing. So we're safe?
No. It means none of its checks fired. It doesn't prove something is safe.
How much is it?
Sell it as the Agent Policy Red Team: $1,500 standard, $3,500 fleet, 72 hours.
One-page summary (paste into a document or email)
Policy Red-Team by FIRME AI CORPORATION, a California corporation What it does Policy Red-Team attacks a company's AI agent rules with known bypass tricks, writes the smallest fix, and proves it with a second independent implementation. The evidence We recorded 5 live checks, including 5 attack cases it had to refuse or flag, and every one matched. You can re-run them yourself without an account. Recorded 2026-11-03. 5 live checks. Median round trip 56 ms including network; this is not a load test or an SLA. Check it yourself https://cainstudio.online/ihvrk-policy-redteam.html Zip SHA-256: 4e3c2424a1b90cfe916916a5b9cd00b2ff70c4d10f7fed116467a863558c735f Signing key fingerprint: 1cacdfc1fc2024b6bb4f8fb02c7c80f4a57f44c5adb11c507190781b5d613de3 Run: python3 verify.py --live (Python 3, no account, no code of ours) What it does not do Written down here: https://cainstudio.online/top25-policy-redteam-limits.html What this is not Not a third-party audit, a certification or a guarantee. No SOC 2 or ISO certification is held. Contact [your name], working with FIRME AI CORPORATION, [your email]
After the call: the proposal email
Hi [first name], Thanks for your time. As discussed, here is what I suggest for Policy Red-Team. What you told me: [their problem, in their words] What we'd do: Sell it as the Agent Policy Red Team: $1,500 standard, $3,500 fleet, 72 hours. Evidence you can check before deciding: https://cainstudio.online/ihvrk-policy-redteam.html Answers for your security review: https://cainstudio.online/evidence-policy-redteam-answers.html Figures for your files: https://cainstudio.online/evidence-policy-redteam-figures.html Limits: Written down here: https://cainstudio.online/top25-policy-redteam-limits.html Next step: [date], [who]. The founder, Vincent Noriega, handles contract and setup. [your name], working with FIRME AI CORPORATION [your postal address] If you'd rather not hear from me again, just say so and I won't write again.
Hand-off to the founder
To: vincent@firmeai.com Subject: Lead: Policy Red-Team - [company] Company: [company] Person: [name], [job title], [email] What they said: [paste their reply] Evidence already sent: https://cainstudio.online/ihvrk-policy-redteam.html Price mentioned: [what you quoted, or none] Next date: [date]
The rules
- Send each message yourself, one person at a time. No bulk lists and no automated sending.
- Commercial email needs your real name, a real postal address and a way to opt out, and you stop when someone asks (US CAN-SPAM; many countries are stricter, so check yours).
- Never change a number, invent a customer, or say certified, audited or guaranteed.
- Only test a system the client owns or operates, after they confirm it in writing.
- Your share is agreed in writing with FIRME AI CORPORATION before your first message (email vincent@firmeai.com). Nothing here promises income. Most messages get no reply; that is normal.