Agent Readiness Audit
Coding agents fail in predictable places: files too big to hold in context, missing docs, no tests to check their own work, no project guidance. The Agent Readiness Audit scores one repository against six weighted pillars, then hands you a prioritized roadmap and one honest number.
$2,500 per repository. The $300 call fee is credited toward your audit.
Documentation and agent-specific readiness carry half the score between them, because they decide whether an agent starts a task oriented or lost.
25% of score
README, architecture docs, and ADRs an agent can learn from without reading every file.
25% of score
CLAUDE.md, nested guidance, skills, and IDE configs an agent can actually use, not decorative files that only look right.
15% of score
Linting that is enforced, a CI pipeline, pre-commit hooks, code review, and honest secret hygiene.
15% of score
Whether agents have a feedback loop to check their own work, or a human is the only one who can.
10% of score
God files and average file size, weighted by whether agents actually need to touch them.
10% of score
Whether code is findable by natural-language search, judged against your language's own conventions.
One composite score from 0 to 100, mapped to a plain-English reality about how agents behave in your repo.
0 to 20, Agents Are Lost. Agents hallucinate and cannot navigate. Every task needs a human driving.
21 to 40, Agents Struggle. Heavy supervision; you correct most of what they produce.
41 to 60, Agents Survive. Useful but slow, with frequent corrections.
61 to 75, Agents Can Work Here. Productive on routine tasks on their own.
76 to 85, Agents Are Productive. Independent on most tasks; supervision becomes review, not hand-holding.
86 to 100, Agents Thrive. A force multiplier across the codebase.
Most codebases that have never been tuned for agents land between 40 and 65. That is not a failing grade. It means the foundations are there and the gaps are fixable, which is exactly what the roadmap is for.
Top features
One client in flight per two-week rotation, a hard cap. That keeps the audit deep instead of assembly-line. Read the full methodology and score bands at agentreadiness.ai.
No. The default engagement is read-only and I do not keep a copy after delivery. If source access is a blocker, the score can run on extracted metadata alone: file counts, structure, config presence, and paths. NDA-friendly either way.
No. A beautifully written codebase can still score low if an agent cannot navigate it. The question is narrow: how well can an AI agent work in this repo right now. If I find a real exposed secret I tell you plainly, but I do not pad the report with security theater.
The audit itself is fast. Scheduling depends on the rotation: I take one client in flight per two-week window so the work stays deep. Book a call and I will tell you the next open slot.
You get the findings and a prioritized roadmap you can act on. I never delete tracked files, never generate files just to move the number, and never recommend key rotation as busywork. Every tool I suggest is real and verified.
Book a short call. The $300 fee comes off your audit, and you leave with a next step either way.