Cursed Agentic

Agent Stack Setup

Claude Code, Cursor, or Codex set up on your repo so agents ship code under review

I add project rules, stop-and-ask boundaries, secret handling, a run-cost view, and one repeatable task with a reviewable diff. I leave a recording and a runbook.

Standard: $500. Founding price for the first 3 jobs. Five business days after I can push to your repo.

I am Ryan Doubrava. I have 14 years in software engineering. My public work: 147 PRs across 11 repos in 39 days.

Start Standard ($500) Book a free 15-minute harness check

GitHub is the usual VCS. GitLab, Bitbucket, or SVN is fine. Email ryan@cursedagentic.com if you cannot book.

If working access never happens, you get a full refund. On Standard, if the day-1 workflow does not leave a clean, reviewable diff on your repo, full refund.

Which tier?

Standard: $500. Founding price for the first 3 jobs. The Token Triage burn audit is a separate engagement.

Harness Audit

$250

I review your repo and rules. You get a gap report and starter rules.

  • Async review of your repo and current rules
  • Written gap report: what agents will break, what to fix first
  • Starter rules you can keep even if you stop there
  • Turnaround: 3 business days after access works
  • Fix window: 1 light fix (48h)
Checkout Harness Audit

Team gates

$1,500

I add routing notes, GitHub review wiring for one repo, an adversarial review pass, and a CI gate sketch.

  • Everything in Standard
  • Routing notes: which agent handles what, with a fallback
  • Local-model fallback: one harness, one model, hardware permitting. Feasibility confirmed at kickoff
  • No-retain routing for your model provider, written down
  • GitHub agentic review on one repo, one review workflow, using your API key
  • An adversarial review pass on one sample change
  • CI gate sketch that fails closed on dangerous agent paths
  • Turnaround: 8 business days after the kickoff call
  • 2 fix passes (7 days) + 30-min live Q&A
Checkout Team gates

If working access never happens, you get a full refund. On Standard, if the day-1 workflow does not leave a clean, reviewable diff on your repo, full refund.

How does it work?

  1. Pay, or book a free 15-minute harness check first.
  2. Send intake and working repo access.
  3. I deliver Standard five business days after I can push to your repo.
  4. I send a recording and a runbook, with one fix pass.

What is out of scope?

I do not build your product or take over your app.

FAQ

What is an agent harness?

An agent harness is the app you type in, plus the project rules and workflow around it. Cursor, Claude Code, and Codex are harnesses. The install is what makes agents ship reviewable code instead of thrashing the tree.

What is Agent Stack Setup?

Agent Stack Setup is a fixed-price install of Claude Code, Cursor, or Codex on your repo so agents ship code under review.

What do I get with Harness Audit?

I review your repo and rules. You get a gap report and starter rules.

What do I get for $500?

I install project rules and one repeatable workflow, with burn visibility and a handoff. Standard includes project rules, secret handling, token-burn triage, one repeatable day-1 workflow that leaves a clean diff, a recorded handoff, and a written runbook. I document how to use secrets without pasting them into chat. One fix pass after handoff. Founding price for the first 3 jobs.

What do I get with Team gates?

I add routing notes, GitHub review wiring for one repo, an adversarial review pass, and a CI gate sketch.

How long does it take and when does the clock start?

I deliver Standard five business days after I can push to your repo. Payment alone does not start the clock. Harness Audit is 3 business days after access works. Team gates is 8 business days after the kickoff call. Org approval delays pause the start.

Do you need my API keys?

Prefer collaborator access or a short-lived fine-grained token scoped to one repo. Never paste long-lived secrets into chat.

What if my repo is a mess?

I will say so early. I either point you at Harness Audit or pause until there is a safe surface.

Do I get your private tools' source?

No. Adapted patterns and configs in your tree. Private Cursed_* repos stay private.

What is the refund policy?

If working access never happens, you get a full refund. On Standard, if the day-1 workflow does not leave a clean, reviewable diff on your repo, full refund.

Can this run on a team monorepo?

I can work on a team monorepo with a clear in-scope package.

Will you build my app?

No. This sets up the stack you use to build. Product work is a separate engagement.

Which harness do you set up: Cursor, Claude Code, or something else?

I configure project rules and a repeatable workflow around the tools you already use. Cursor, Claude Code, Codex, Hermes, OpenCode, Continue, Cline, Antigravity, or similar. More than one if already installed and called out in intake. Say which is primary.

Is this AI automation / n8n / Zapier?

No. Engineering agent stack inside your IDE/CLI against a real repo.

Is Harness Audit the Token Triage burn audit?

No. Harness Audit reviews your repo and rules before an install. The Token Triage burn audit is a 4-6 hour read of what your agents actually spend, for teams on API keys. It has its own page: https://cursedagentic.com/burn-audit/

Intake

Book the kickoff and answer five questions. Email ryan@cursedagentic.com as a fallback.

This form previews an email. I receive your answers only after you send the email. Do not send API keys.

Book the kickoff and answer five questions