Smakosh®

Get your team shipping with AI agents, safely.

I set up the skills, guardrails and review gates that turn coding agents from a lottery into a repeatable way of shipping, then train your team to run it.
499
merged pull requests shipped with agents on LLM Gateway
9
repo-local skills running in production there
1 key
across Claude Code, Cursor, OpenCode and Copilot with DevPass
SOC 2
Type II program I led while shipping this way

01What you get

The outcome, not the hours.

Who it is forEngineering leaders whose team has Copilot, Cursor or Claude Code licences and uneven results to show for them.

  1. 01

    Agent output your seniors will merge

    Conventions written down as skills and checked by tests and review gates, so quality stops depending on who wrote the prompt.

  2. 02

    A workflow the whole team follows

    The same setup in every repo: skills, MCP servers, sandboxes and a pull request routine everyone understands.

  3. 03

    Spend and risk under control

    One gateway for model access, budgets per team, and clear rules for what code and data an agent may touch.

  4. 04

    People who know what good looks like

    Hands-on training on your code, including when not to use an agent at all.

02Pricing

Fixed prices, written scope.

No hourly meter and no surprise invoice. You see the price before the work starts, and it does not move once it has.

  1. 011 week

    AI readiness audit

    A clear picture of where agents will pay off in your codebase, and what is in the way.

    • Interviews with leads and engineers
    • Repo, CI and review process assessment
    • Tooling, cost and data-policy review
    • Written roadmap and a 60 minute readout

    $2,900

    fixed price

    Book this package
  2. 023 weeks

    Team rollout

    Where most projects start

    Agents set up properly in your repos, and your team trained to work with them.

    • Skills, MCP servers and sandboxes in up to three repos
    • Review gates and context hygiene configured
    • Half-day hands-on workshop for the team
    • Before and after metrics on a pilot project

    $9,500

    fixed price

    Book this package
  3. 03monthly, 3 month minimum

    Adoption partner

    A standing senior voice as models, tools and your codebase keep changing.

    • Two working days a month
    • Skill and routine maintenance
    • Office hours for engineers
    • Monthly review of usage, cost and quality

    $4,500

    per month

    Book this package

Prices in USD, excluding VAT where it applies. The audit price comes off the rollout if you book it within 30 days. Model and tool licences are yours and billed by their vendors.

03How it runs

Four steps, no mystery.

  1. Step 01

    Audit

    How the team works today: tools, repos, review habits, security constraints and where agents already help or hurt.

  2. Step 02

    Pilot

    One repo set up properly with skills, guardrails and a measured before and after.

  3. Step 03

    Rollout

    The pilot setup templated across repos, with a workshop for every engineer who will use it.

  4. Step 04

    Sustain

    Scheduled routines, a skill that audits the other skills, and a monthly review so the setup does not go stale.

04Background

The long version.

Most teams I talk to are in the same place. Everyone has an AI coding tool, a few people get a lot out of it, most get a little, and nobody can say whether the pull requests got better or just bigger.

The gap is rarely the model. It is everything around it.

#What I have learned doing this for real

I build LLM Gateway with coding agents every day, and the setup that makes their output trustworthy lives in the repo next to the code:

  • Repo-local skills that encode the conventions an agent would otherwise guess at: how to add a provider, write a migration, open a pull request, cut a changelog entry. They are reviewed like code.
  • A skill that audits the other skills, because agent instructions fail silently by going stale, not loudly by breaking.
  • Scheduled routines that run recurring work unattended and open pull requests with fixes, instead of reports nobody reads.
  • An isolated stack per worktree, with its own ports, Postgres and Redis, so several agents work in parallel without colliding.
  • Deliberate context hygiene. Deny-lists keep build output and generated files out of the context window. This does more for quality than switching models.

At Axis Communications, where I consult through Sigma Technology Group, I have shared the same workflows with the team: GitHub Copilot, reusable skills, MCP integrations and agent orchestration.

I also build the tools on the other side of this. DevPass gives a team one key and a flat price across models in Claude Code, OpenCode, Cursor and Copilot, and DevPass Code is our own open-source terminal coding agent.

#What you typically hire me for

  • Finding out, honestly, whether agents are helping your team today and where they would help most.
  • Setting up skills, MCP servers, sandboxes and review gates in your repositories.
  • Training engineers on your own codebase, not on a toy app.
  • Choosing tools and model access without getting locked into one vendor.
  • Writing the policy for what agents may and may not touch, in language engineers will follow.

If your team has the licences but not the results, tell me where you are and I will tell you what I would do first.

05Questions

Asked before signing.

We already pay for Copilot. Why do we need this?

A licence gives you autocomplete. The gains come from the scaffolding around it: written conventions, tests, review gates and isolated environments. That is the part most teams never set up.

Which tools do you work with?

Claude Code, Cursor, GitHub Copilot, OpenCode and Codex, plus MCP servers and the Vercel AI SDK. The setup is written to survive a change of tool.

Is our code safe?

We agree the data rules first: which providers, which repos, what stays out of context. I have led a SOC 2 Type II program and set this up accordingly, under NDA.

How do you measure whether it worked?

On the pilot we pick measures before starting, usually cycle time, review rounds and defects on comparable work, and report them after. No measure, no claim.

Will this replace engineers?

No. Agents are fast at producing plausible code, and plausible is the failure mode that costs a quarter. This makes good engineers faster and keeps a human reading every diff.

Tell me what you are building

A first call is 30 minutes and free. You leave with a view on scope and price whether or not we work together.

Book a scope call