Context fluency is a term proposed in Mise en Place for Agentic Coding: Deliberate Preparation as Context Engineering Methodology for the ability to create rich, structured context that AI agents can act on. The author presents it as a skill distinct from both traditional programming and prompt engineering: it captures domain expertise, value judgments and design intent in machine-legible form, not as one-shot prompts but as persistent informational environments. Where Prompt engineering optimizes individual instructions, context fluency is described as upstream of prompting, concerned with the informational architecture surrounding agent execution; and where the paper’s Mise en Place (Agentic Coding Methodology) describes a sequence of phases producing artifacts, context fluency names the practitioner skill that makes that process effective.
Usage
The paper identifies four components of the skill. Decomposition is breaking problems into discrete, parallelizable tasks with clear boundaries so agents can execute concurrently. Specification is describing not only what to build but why, so agents can make aligned micro-decisions without human intervention. Constraint definition is knowing what to exclude, simplify or defer, treating scope management as a first-class concern. Domain encoding is externalizing tacit knowledge that agents cannot generate on their own; in the paper’s hackathon case, twenty minutes of dictated pedagogical intuitions are reported to have substantially reduced iteration on the domain-specific tutor component.
The author connects these components to backward design (outcome-driven preparation), to Polanyi’s account of tacit knowledge — “we know more than we can tell”, with context fluency described as the discipline of telling it anyway — and to pedagogical scaffolding, structuring an environment so that a learner, or an agent, can act independently. If it is a distinct skill, the paper suggests, developer education should cultivate specification, decomposition and domain encoding alongside programming; practitioners with strong domain knowledge and pedagogical instincts may be disproportionately effective in agentic workflows; and preparation aids such as task systems, specification frameworks and context-engineering platforms become essential infrastructure. Whether context fluency is trainable or a function of prior domain knowledge is one of the open questions the paper poses, and it has not been empirically validated.