
The emergence of large language model (LLM)-based coding agents has fundamentally altered the pace and nature of software production. Technical friction — the time required to translate intent into working code — has been substantially reduced. This transition introduces both a significant opportunity and a commensurate risk: the same capability that accelerates high-quality software production also accelerates the production of low-quality software at scale. We propose VibeOps, a governance and optimisation framework for agentic coding environments. VibeOps introduces four primary contributions: (1) a structured environment contract (AGENTS.md) that governs agent behaviour through version-controlled, team-owned configuration; (2) a portable methodology layer implemented as Agent Skills; (3) VibeEvals, a multi-dimensional evaluation framework that treats development sessions as controlled experiments with separable parameters and metrics; and (4) the 12 Vibing Factors, an opinionated set of design principles for governed AI-assisted development. We further introduce Architecture Distance as a novel metric for measuring the alignment of agent-generatedcode to a team's canonical blueprint, and Topic Trajectory Analysis as a method for extracting behavioural signals from structured session logs. We argue that governing the agentic coding environment — rather than optimising the agent or the prompt — is the highest-leverage intervention available to engineering teams, and that an experimental approach to environment construction is both necessary and sufficient for systematic improvement.
Artificial intelligence, Machine learning, Software development, Machine Learning/standards
Artificial intelligence, Machine learning, Software development, Machine Learning/standards
| selected citations These citations are derived from selected sources. This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically). | 0 | |
| popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network. | Average | |
| influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically). | Average | |
| impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network. | Average |
