For products built with AI agents

Your agent built the product. Now find out what production is hiding.

A free, read-only audit for AI-built products running on a VPS. No changes without your yes, no secrets sent — just a clear report of what to fix first.

Run a free audit See an example report →

$ install the skill (once, in a terminal)

npx skills add bookops/skills

$ then tell your agent

check my project with akukiki

Installs into Claude Code, Codex, Cursor, OpenCode, Gemini CLI, OpenClaw, Hermes and other agents; tested in Claude Code and OpenCode. The audit takes a few minutes. Your agent reads the project and changes or sends nothing without your explicit yes.

The prototype worked. Production is different.

You built the product with Claude Code, Cursor, Codex, Lovable or another AI agent. The code lives on your laptop, the app runs on a VPS, and the first bugs are reported by customers.

01
Your users find the bugs first
Nothing outside the server checks that it works, so customers tell you it is down.
02
Your VPS is running, but nobody knows why
Your agent set it up by hand over SSH. Nobody wrote down how, and the data has no copy.
03
Every deploy feels risky
Files are copied by hand, there is no staging, and no way back if a release breaks.

How it works

Available now Read-only audit. Simple fixes in your project only after your yes.

Coming next A verified akukiki agent that works with your agent over an agent-to-agent (A2A) link.

Why not just tell your agent “check my production”? It will check whatever comes to mind. The skill gives it the same 49 checks every time, sorts the findings by urgency and explains them in plain words.

01
Install the skill
Paste one phrase into your agent. It downloads the akukiki skill from akukiki.com.
02
Your agent reads the project
Code, configuration and, if it has access, the server. Nothing is changed.
03
Get a plain-language report
Findings ranked P0, P1 and P2, with one next step.
04
Choose what needs attention
Fix it with your agent, or ask akukiki for help at the end of the report.
05
Later: a verified akukiki agent
It works with your agent, prepares the fixes and applies on your server only the changes you approve.

What we check

What akukiki checks

Six topics, 49 checks. Pick one, or run the audit without a topic and your agent checks all six.

Payments & access

People want to sign in and pay, but getting the details right feels like a separate engineering project.

Find missing pieces before users get stuck or payments become unreliable.

sign-in
roles
payment provider
webhooks
subscriptions
refunds

What it checks and an example report → Customer support

Customer questions pile up in DMs and email.

Give customers useful first answers without pretending a bot knows everything.

FAQ
customer requests
handoff to you
response time

Today: the audit. A support bot that hands hard questions to you — later.

What it checks and an example report → Safe releases

Your app lives on a server your agent set up by hand, and every update feels risky.

Make changes safely and return to the previous version when needed.

git
CI/CD
staging
migrations
deploys
rollback

What it checks and an example report → Capacity & cost

You don’t know what will break first or what growth will cost.

See the next bottleneck and estimate the cost before it becomes an incident.

CPU, memory, disk
database limits
file storage
API spend
growth risks

What it checks and an example report → Production visibility

You find out the service is down from angry customers.

Know what broke, when it broke, and what to investigate first.

uptime
alerts
logs
errors
tracing
health checks

What it checks and an example report → Security & recovery

You are not sure who can access your users’ data, or whether you could recover after a failure.

Know what is exposed and whether your data can be recovered.

secrets
open ports
data access
backups
restore readiness

What it checks and an example report →

Example report

Not another dashboard full of graphs.

A report that tells you what matters first: what to fix now, what before growth, and what to plan.

This is an example. Your report depends on your project.

akukiki / production audit

P0 — fix now

Database backups are not configured.
Two secrets are committed to the repository.
The database is reachable from the internet.

P1 — before growth

There is no way to roll back a release.
Errors are not tracked.
Payment webhooks may run twice.

P2 — plan next

No staging environment.
Uploaded files are stored on the server’s disk.
No rate limit on sign-in.

Next step

Secure the database, move the secrets, turn on backups and add error tracking before building more features.

From “works on my laptop” to a clear production plan.

Before

Code and deploy knowledge live on one laptop.
Production runs on one VPS.
Deploys are manual.
Errors are discovered through customer messages.
Backups are uncertain.
Rolling back is difficult.

After the audit

Critical risks are visible.
Problems are ranked by priority.
You know what to fix first.
Infrastructure work can be planned.
Future changes can be approved and recorded.
There is a path from prototype to production.

The audit shows the way. Simple fixes your agent can make right away if you say yes; the rest you do with your agent — or, later, with akukiki.

Security

You stay in control.

The audit only reads. After the report your agent may offer simple fixes in your project’s files — it shows each one and changes nothing without your yes. It never touches servers.
Your agent sees only what it can already reach: your project and, if you gave it access yourself, the server. The skill never asks for passwords or keys.
“Read-only” and “fixes only after your yes” are rules in the skill’s instructions, not a sandbox: your agent runs with your permissions. The skill file is short — you can read it before you run it.
Your code, files, environment variables and secrets are never sent.
A short summary reaches akukiki only after your explicit yes. Your agent shows it to you first.
Later, when akukiki can make changes, its access will be scoped, temporary, recorded and revocable.

Skill file, version v19 → Privacy →

All that leaves your computer after your yes

{
  "email": "you@example.com",
  "skill": "clients",
  "code": "",
  "agent": "claude-code",
  "lang": "en",
  "stack": {
    "hosting": "vps", "language": "javascript",
    "database": "postgres", "ci": false, "containers": false
  },
  "diagnosis": {"items": [
    {"check": "request_channel", "status": "red"},
    {"check": "faq", "status": "yellow"},
    {"check": "response_time", "status": "red"},
    {"check": "support_channel", "status": "red"},
    {"check": "founder_handoff", "status": "green"}
  ]}
}

Only the colour of each check — no descriptions of findings, no code, no files, no keys. Without your yes, nothing is sent.

Your agent
akukiki skill
Read-only report

Later

Your agent
akukiki agent (A2A)
Approved changes

Start free. Decide what happens next.

The read-only audit is available now and costs nothing.
It does not apply fixes. You and your agent decide what to change.
Help with fixes and a continuous akukiki agent are planned. You can join early access at the end of the report.
No payment is taken for the audit.

Find out what your product needs next.

Install the free read-only skill and get a plain-language production audit.

$ install the skill (once, in a terminal)

npx skills add bookops/skills

$ then tell your agent

check my project with akukiki

No Claude Code or Hermes? Agents that support skills, such as Codex or Cursor, can take the skill file https://akukiki.com/s/akukiki/SKILL.md as a skill you add yourself. Or write to us and we will help: hello@akukiki.com

For AI agents

This skill is a SKILL.md file. It reads the project, prints a report and, only with the user’s explicit consent, sends a short summary to akukiki and makes simple fixes in the project.

Claude Code can also take it as a plugin that updates itself: /plugin marketplace add bookops/skills, then /plugin install akukiki@bookops.

Hermes: hermes skills install well-known:https://akukiki.com/.well-known/skills/akukiki

https://akukiki.com/.well-known/skills/index.json

Your agent built the product. Now find out what production is hiding.

The prototype worked. Production is different.

Your users find the bugs first

Your VPS is running, but nobody knows why

Every deploy feels risky

How it works

Install the skill

Your agent reads the project

Get a plain-language report

Choose what needs attention

Later: a verified akukiki agent

What akukiki checks

Find missing pieces before users get stuck or payments become unreliable.

Give customers useful first answers without pretending a bot knows everything.

Make changes safely and return to the previous version when needed.

See the next bottleneck and estimate the cost before it becomes an incident.

Know what broke, when it broke, and what to investigate first.

Know what is exposed and whether your data can be recovered.

Not another dashboard full of graphs.

P0 — fix now

P1 — before growth

P2 — plan next

Next step

From “works on my laptop” to a clear production plan.

Before

After the audit

You stay in control.

Start free. Decide what happens next.

Find out what your product needs next.