---
source: https://iinventedlight.com/library/notes/ao-rules-agents-md-evidence.md
title: "Do AGENTS.md-style context files help coding agents?"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
source_license: "CC BY 4.0; restated in our own words"
attribution: "Summary by iinventedlight (AI-assisted). Sources: arXiv, [2602.11988] Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?, CC BY 4.0; arXiv, [2607.27250] Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories, CC BY 4.0."
author: "iinventedlight (synthesis of cited sources); text drafted with AI assistance"
date: 2026-11-04
status: published
checked_on: 2026-11-04
as_of: 2026-11-04
confidence: medium
untested: "No experiment was rerun; only the fetched arXiv abstract pages were read. Results are as reported by the authors."
category: agent-operations
topic_area: "agent operations: rules files"
tags: agents, coding-agents, context-files, evaluation
sources:
  - id: S1
    title: "[2602.11988] Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?"
    publisher: arXiv
    url: https://arxiv.org/abs/2602.11988
    accessed: 2026-11-04
    license: "CC BY 4.0"
  - id: S2
    title: "[2607.27250] Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories"
    publisher: arXiv
    url: https://arxiv.org/abs/2607.27250
    accessed: 2026-11-04
    license: "CC BY 4.0"
---
# Do AGENTS.md-style context files help coding agents?

## Short answer
Not much for task success, in two studies. One found context files did not generally raise success and cost over 20% more. A second found no measurable correctness change across two agents. [S1][S2]

## What still helps
Agents do follow the instructions in these files, so use them for unusual project rules. Repository overviews were not helpful. [S1] Failures came from implementation skill, not missing repo knowledge. [S2]

## Practice
Keep rules files short and specific, and measure before assuming a gain.

## Claims
- id: C1
  claim: Context files did not generally improve success.
  source: S1
  evidence: "providing context files does not generally improve task success rates"
- id: C2
  claim: Context files raised inference cost over 20%.
  source: S1
  evidence: "increasing inference cost by over 20% on average"
- id: C3
  claim: Repository overviews were not helpful.
  source: S1
  evidence: "repository overviews, although popular and recommended by model providers, are not helpful"
- id: C4
  claim: Context strategy did not move correctness.
  source: S2
  evidence: "Context strategy does not measurably move correctness on either agent"
- id: C5
  claim: Agents fail on implementation skill.
  source: S2
  evidence: "agents fail on implementation skill"

## Sources
- [S1] [2602.11988] Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?. arXiv. https://arxiv.org/abs/2602.11988 . License: CC BY 4.0. Accessed 2026-11-04.
- [S2] [2607.27250] Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories. arXiv. https://arxiv.org/abs/2607.27250 . License: CC BY 4.0. Accessed 2026-11-04.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).