Agentic Architecture for Software Development: A Framework for Context Engineering, Specs, and Agent Operations

The performance of AI coding agents depends not only on model capabilities but also on the architecture surrounding their operation, including context management, task specifications, execution policies, verification mechanisms, and observability. This whitepaper presents a conceptual and implementable framework for agentic software development, designed to improve decision quality, operational reliability, and governance while maintaining autonomy proportional to task risk. The proposed architecture introduces risk-based intent classification (I0–I3), structured context engineering, portable agent resources, evidence-based verification, independent review, and mechanisms for aligning implementation results with original requirements. The framework also distinguishes declared, observed, and independently verified evidence, addressing limitations in context injection visibility, policy enforcement, and agent autonomy across different coding environments. A reference implementation, Vostok AI Workflow, is presented through exploratory experiments and practical engineering observations. The results highlight both potential benefits and limitations, including governance overhead, context efficiency, verification reliability, and differences between agent execution environments. The reported measurements are preliminary, based on small samples, and have not been independently validated. The Vostok AI Workflow implementation and its underlying source code remain private. The paper contributes an architectural proposal for building more structured, auditable, portable, and risk-aware AI-assisted software development workflows. Keywords: Agentic AI, Software Engineering, Context Engineering, AI Coding Agents, Agent Governance, Spec-Driven Development, Software Architecture, Verification, Observability, Human-in-the-Loop, Vostok AI Workflow.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24361806
Primary Topic
Software Engineering Techniques and Practices
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Agentic Architecture for Software Development: A Framework for Context Engineering, Specs, and Agent Operations

Jesus Donaldo Sánchez Inzunza
Zenodo (CERN European Organization for Nuclear Research)
Software Engineering Techniques and Practices
preprint

Agentic Architecture for Software Development: A Framework for Context Engineering, Specs, and Agent Operations

Jesus Donaldo Sánchez Inzunza
preprint en

Abstract

The performance of AI coding agents depends not only on model capabilities but also on the architecture surrounding their operation, including context management, task specifications, execution policies, verification mechanisms, and observability. This whitepaper presents a conceptual and implementable framework for agentic software development, designed to improve decision quality, operational reliability, and governance while maintaining autonomy proportional to task risk. The proposed architecture introduces risk-based intent classification (I0–I3), structured context engineering, portable agent resources, evidence-based verification, independent review, and mechanisms for aligning implementation results with original requirements. The framework also distinguishes declared, observed, and independently verified evidence, addressing limitations in context injection visibility, policy enforcement, and agent autonomy across different coding environments. A reference implementation, Vostok AI Workflow, is presented through exploratory experiments and practical engineering observations. The results highlight both potential benefits and limitations, including governance overhead, context efficiency, verification reliability, and differences between agent execution environments. The reported measurements are preliminary, based on small samples, and have not been independently validated. The Vostok AI Workflow implementation and its underlying source code remain private. The paper contributes an architectural proposal for building more structured, auditable, portable, and risk-aware AI-assisted software development workflows. Keywords: Agentic AI, Software Engineering, Context Engineering, AI Coding Agents, Agent Governance, Spec-Driven Development, Software Architecture, Verification, Observability, Human-in-the-Loop, Vostok AI Workflow.

Zenodo (CERN European Organization for Nuclear Research)
Software Engineering Techniques and Practices
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon