Is your codebase ready for coding agents?
Agents can write more code in an afternoon than your team can review in a week, and without human eyes on it, there's no reliable way to know that it's right.
LLMs find shortcuts to deliver to the spec they're given: casts to quiet the compiler errors; guards that turn crashes into superficially correct code giving subtly wrong results.
Codebases used to evolve through human-scale processes: developers crafting code and reviewers reading every line. With agents flooding the pipeline, these become choke points - a team is overwhelmed by the volume of code. Reviews become rubber-stamping or agent-only, and code unread by humans silently accumulates in the repo, with nobody able to answer for it.
Code reviews that carry weight
The same team and repo: your reviewer sits down to judge whether a code diff matches the intent of the spec. Review is about design and product decisions. Approval means the reviewer puts their name to the code going into the repo.
They see that agent shortcuts from previous diffs are blocked. The same trick can't work twice. The number of ways the code can be wrong decreases. Feature work never stopped.
Book a free 30-minute agent-readiness check
You'll walk us through your codebase and how your team uses agents today. If you give us read-only access in advance, we can run our tools against it and bring the output to the call. They're most advanced for TypeScript, and we're actively developing PHP and WordPress ones. By the end you'll know which checks would help the most and whether an agent-readiness review is worth doing. You'll hear back from us within one business day.
In the agent-readiness review we run checks against your whole codebase and provide a written report with numbers, and recommendations for the highest-value checks you can put in place. We provide the configuration for our open-source tools so you can run it yourself and watch the numbers go down.
These are checks we run on our own code and our clients' codebases every day.
"Ed and his team at Autotelic are an agency that we keep doing work with as they are comfortable with ambiguity and continue to deliver really great solutions and excellent work for us."

Do we even need this if we already have AGENTS.md and skills?
To paraphrase Captain Barbossa, those are more what you'd call guidelines than rules. Guidelines can be ignored or lost in context, but a failing check can't be ignored. Deterministic checks are complementary to your existing setup.
Can't agents do the reviews?
Yes, but those LLMs also can't be trusted. Models that review share the training and blind spots, and verdicts vary between runs so they can't settle on "done". Deterministic checks constrain what can be merged and ensure code conforms to the rules you set.
We can't afford to slow down feature work
Our approach isn't a code freeze blocking work until everything's fixed. We have a ratchet for existing code: today's findings are accepted as baseline, and fixed as files are worked on. As classes of errors are reduced to zero, the ratchet is turned so the code can't regress.
Is this relevant if our stack isn't TypeScript?
The approaches are language-agnostic and we'd love to do a call about any language and stack, but would probably have a different initial conversation - please do contact us to discuss.
How much does this cost?
The 30-minute agent-readiness check is free. The agent-readiness review is priced per codebase and agreed before we start, and is delivered within 2 weeks. If we're not the right fit we'll say so, but if it's something you could use we'd love to work with you.