- Home
- Skills
- Agents & Orchestration
- KERNEL Gate Pro
Works with the AI Agents you already use
KERNEL Gate Pro
The KERNEL gate with teeth: a scoring script that refuses to bend the rubric, a hard risk veto, fixed triggers and an append-only log, with 17 tests that try to break it.
$9
KERNEL Gate Pro
Example session with this skill installed
We need users to upgrade their subscription from /billing and move legacy-plan users to the new plans. Score the ticket.
- Read your context and instructions
- Compiled the kernel gate pro
I declared the triggers first
$ python3 scripts/kernel.py needed --change build --change migration --change data
KERNEL NEEDED
Then I wrote the ticket and scored it
$ python3 scripts/kernel.py evaluate tickets/billing.kernel.json --log
KERNEL 84/100 · RISK_REVIEW
Weakest: risk 11/15
clarity 17/20: One goal per part and a clear upgrade path; the success metric is not written
scope 18/20: Inclusions and exclusions written; what happens to annual plans is not decided
context 13/15: Billing module, Stripe mode and legacy schema are named; webhook setup is not documented
risk 11/15: Rollback script and idempotent webhooks planned; no plan yet for a payment that fails halfway
validation 13/15: Webhook replay and a dry-run migration on a copy of the data; no test for the failed-payment case
priority 12/15: Unblocks the new pricing launch; no date attached
Risk veto: migration, payments → a second reviewer reads the plan first
Logged in .kernel/kernel.jsonl
84 passes the line, but this ticket touches payments and a migration, so it goes to review anyway (exit code 3). I will not write code until someone other than me has read the plan.
If anyone tries to round a score up, the script refuses:
KERNEL REFUSES: risk: whole number between 0 and 15
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
KERNEL Gate Pro scores a ticket before your agent builds anything: six weighted dimensions, 100 points, a hard line at 80. The free KERNEL Gate tells the agent how to score. Pro makes the score hard to fake.
What Pro adds
- Triggers. A fixed list of changes that always need the gate (build, data, migration, permissions, publish...). It never depends on how the request is phrased.
- A scorer that refuses. A Python script computes the total and the verdict. Invented dimensions, generous decimals, scores without reasons and unknown triggers are rejected.
- A hard risk veto. Migrations, auth, credentials, personal data, payments, tenant isolation and deletion always go to a second reviewer, even at 100/100.
- An append-only log of every evaluation, so you can look back at what was scored and why.
- 17 tests, 12 of them sabotage attempts that must be refused. If one stops failing, the gate is broken.
Exit codes your agent cannot ignore
EXECUTE (80+) exits 0. REVIEW and RISK_REVIEW exit 3. REJECT (under 60) exits 4. An invalid evaluation exits 2. The skill tells the agent never to chain the command in a way that ignores the exit code.
Why this beats prompting it yourself
A prompt can be talked into a better score. A script with fixed weights cannot: a score only goes up when the missing piece is written down. And an average hides the one thing that can sink you; the veto makes it visible every time.
Python 3.8+, no dependencies, no network. Made by Alex Cano at LexLabs; we run this gate on our own work every day.
Requires Python 3.8+.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Requires
Permissions
Python 3.8+