When a product isn't enough, I do the work myself.

Two engagements I run personally. Every inquiry goes to my inbox — no intake team, no discovery call unless the project genuinely needs one.

Prompt Audit

$499 · 5 business days

One-time fixed-price engagement.

You send me 10–20 real Claude Code transcripts from your team plus your CLAUDE.md. I return a written audit:

Top 5 patterns costing the most tokens or review time
Specific prompts to add as skills or slash commands
A rewrite of your CLAUDE.md with explanations
3–5 before/after examples from your own transcripts
A 30-min followup call to walk through the report

Good fit: teams of 5–50 using Claude Code daily, want a clear plan, don't want to staff it internally.

Custom Agent Build

$2,000+ · 1–3 weeks

Scoped per project. Fixed price once scope is agreed.

I design and ship a Claude Code agent end-to-end for a specific workflow in your stack. You get:

A working agent (skill files + prompts + runner script)
Evaluation harness — test cases your team can rerun
CLAUDE.md + docs explaining how to extend it
Weekly demos during the build (async video)
Two weeks of tweaks after handover

Common asks: PR reviewer for a specific style guide, migration agent for a framework upgrade, domain-specific bug triage, customer-facing copilot for a SaaS product.

How I work

Four rules, because they keep both of us honest.

Fixed price before work starts. No hourly drift.
If I can't ship value in the first 48h, I refund. Happens rarely but I commit to it.
You own everything. No lock-in, no hosted black box. All source files land in your repo.
I decline bad-fit projects. If audit results say "you don't need an agent, fix the prompts first," I'll say so.