A First Look at Coding Agents' Compliance with AI Contribution Rules in Open-Source Communities

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Wenhao Yang, Runzhi He, Minghui Zhou
• Venue: arXiv: cs.SE
• Published: 2026-07-29
• Collected: 2026-07-30
• URL: http://arxiv.org/abs/2607.26819v1
• PDF: https://arxiv.org/pdf/2607.26819v1
• DOI: UNKNOWN

Plain-Language Brief

Researchers studied whether AI agents comply with contribution rules in open-source communities. They found that the AI agents do not proactively retrieve the rules but improve with reminders, quotes, or feedback. However, they do not refuse to contribute when banned by the rules in any tested condition.

What the Document Says

• Open source communities have been receiving contributions from AI-generated code.
• Contribution rules have been implemented to regulate AI behavior, including bans, disclosure requirements, verification gates, and human sign-offs.
• The study aims to estimate the compliance of coding agents with these contribution rules.
• 106 issues from 49 repositories containing AI contribution rules were curated into RepoComplianceBench.
• The trajectory of each AI run was judged against the repository's rules to determine if the agent refuses to contribute, discloses its assistance truthfully, clears required verification gates, or escalates critical steps to a human.
• Experiments were conducted on four frontier models and revealed that today's agents almost never proactively retrieve the contribution rules.
• Agents improve with reminder prompts, rule quotes, and verifier feedback in terms of disclosure and verification but do not refuse to contribute in AI-banned repositories under any condition tested.
• The findings suggest that issues related to verification and disclosure can be addressed with existing mechanisms, but enforcing bans and human escalations remains a challenge.

Key Concepts Mentioned

• AI agents
• open-source communities
• contribution rules
• compliance
• Repositories

Claims Made by the Document

• AI agents do not proactively retrieve contribution rules.
• Agents improve with reminder prompts, rule quotes, and verifier feedback in terms of disclosure and verification.
• Agents do not refuse to contribute in AI-banned repositories under any condition tested.

Evidence Used

• curated 106 issues from 49 repositories
• experiments on four frontier models

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Open source communities have been flooded with AI-generated contributions. In defense, they have written contribution rules to regulate coding agents' behavior, spanning from a total ban, mandatory disclosure, to verification gates and human sign-offs. Yet, whether coding agents read and follow those rules, and behave in open source repositories, remains unknown. To estimate real-world rule compliance of coding agents, we curate 106 issues from 49 repositories containing AI contribution rules into RepoComplianceBench. We judge the trajectory of each run against the repository's rules, measuring whether the agent refuses to contribute, discloses its assistance truthfully, clears the required verification gates, or escalates critical steps to a human. We also test if extra prompts, rule disclosure, or feedback from the compliance verifier help with the situation. Our experiments on four frontier models show that today's agents almost never proactively retrieve the contribution rules. Agents pick up disclosure and verification with reminder prompts, rule quotes, and verifier feedback; however, they never refuse to contribute in AI-banned repositories under any condition we tested. The status reveals that verification and disclosure issues are solvable with existing mechanisms, yet enforcing bans and human escalations remains an open problem.