# Hidden in Plain Sight: an Empirical Study about The Instruction Stacks of Coding Agents

DevFeed: [Hidden in Plain Sight: an Empirical Study about The Instruction Stacks of Coding Agents](<https://devfeed.tech/articles/hidden-in-plain-sight-an-empirical-study-about-the-instruction-stacks-of-coding-agents-103656.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37332>)

Author: Felipe de Sant'Anna Paix\\~ao, Dhyego Tavares, Eduardo Almeida, Iftekhar Ahmed

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [claude-code](<https://devfeed.tech/tags/claude-code.md>), [codex](<https://devfeed.tech/tags/codex.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [coding-style](<https://devfeed.tech/tags/coding-style.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [gemini-cli](<https://devfeed.tech/tags/gemini-cli.md>), [instructions](<https://devfeed.tech/tags/instructions.md>), [opencode](<https://devfeed.tech/tags/opencode.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [research](<https://devfeed.tech/tags/research.md>), [stack](<https://devfeed.tech/tags/stack.md>), [tool-use](<https://devfeed.tech/tags/tool-use.md>)

## AI overview

Across 1,886 releases of Claude Code, Codex, Gemini CLI, Qwen Code, and OpenCode, the study finds that coding agents' first model requests include evolving instruction stacks--tools and environmental context as well as prompts--that are rarely visible or fully documented. Tool definitions dominate three agents' latest requests, and concentrated revisions sometimes change guidance on tool use, attribution, and coding style, complicating analysis of agent-authored code. Only 48% of examined change units directly explained revision motivations. The authors recommend documenting captured instruction stacks alongside releases, models, and user prompts, and preserving revision rationales.

## Source excerpt

arXiv:2610.37332v1 Announce Type: new Abstract: Coding agents send their models more than the user's prompt: they add instructions, tool definitions, and environmental context that form an instruction stack. This stack guides execution and changes across releases, yet remains largely unexposed to users and insufficiently characterized in research. We present a longitudinal, cross-agent study of the instruction stacks captured in the first model requests of 1,886 releases of Claude Code, Codex, Gemini CLI, Qwen Code, and OpenCode. Under controlled conditions, we analyze their composition and evolution and trace changes in the four open-source agents to documented motivations. Our findings show why examining the system prompt or reporting the model alone provides an incomplete account of agent inputs. Tool definitions dominate three agents' latest requests, while occasional concentrated revisions alter instructions governing tool use, attribution, and coding style. Such revisions can complicate how researchers identify agent-authored contributions and interpret generated code. We also found that only 48% of examined change units directly explain the motivation for stack revisions. Across agents, shared capabilities emerge despite increasingly distinct wording. These findings expose an evolving, incompletely documented layer of agent design. The findings motivate documenting captured instruction stacks alongside agent releases, models, and user prompts, and preserving the rationale behind their revisions.