Type: url
Original: https://arxiv.org/pdf/2509.06216
Status: generated
Summary
This paper proposes Structured Agentic Software Engineering (SASE), a framework for the Agentic Software Engineering (SE3.0) era that reimagines software engineering’s actors, processes, tools, and artifacts around a duality between SE for Humans and SE for Agents. It introduces two dedicated workbenches (the Agent Command Environment and the Agent Execution Environment), a set of version-controlled artifacts (BriefingScript, LoopScript, MentorScript, Consultation Request Pack, Merge-Readiness Pack, Version Controlled Resolution), five named engineering activities, and a six-level agency/autonomy hierarchy (SE1.0-SE5.0), grounding its argument in a worked example of a developer orchestrating an agent team and closing with a research roadmap for the field.
Generated pages
- Agentic Software Engineering: Foundational Pillars and a Research Roadmap
- Structured Agentic Software Engineering (SASE)
- SE Autonomy Levels (SE1.0–SE5.0)
- Agent Command Environment (ACE)
- Agent Execution Environment (AEE)
- BriefingScript
- LoopScript
- MentorScript
- Consultation Request Pack (CRP)
- Merge-Readiness Pack (MRP)
- Version Controlled Resolution (VCR)
- Briefing Engineering (BriefingEng)
- Agentic Loop Engineering (ALE)
- AI Teammate Mentorship Engineering (ATME)
- Agentic Guidance Engineering (AGE)
- AI Teammate Lifecycle Engineering (ATLE)
- AI Teammate Infrastructure Engineering (ATIE)
- Plan-Do-Assess-Review (PDAR)
- Product Requirement Prompt (PRP)
- N-Version Programming
- SWE-Bench Verified
- SWE-Bench Pro
- BMAD (Breakthrough Method for Agile AI-Driven Development)
- Devin
- SWE-Bench
- AIDev
- Kiro
- Claude Code
- Google Jules
- GitHub Copilot
- AGENTS.md