Skip to content

llm-contribution-policy.md

Latest commit

 

History

History
112 lines (68 loc) · 8.1 KB

llm-contribution-policy.md

File metadata and controls

112 lines (68 loc) · 8.1 KB
Raw

Algorithmic Collaboration and Contribution Guidelines

Version: 3.0.0 • Last Review: 2026-06-01

Author: Güneş Şevval Arjinya Melisoy AKA Ashley Íris Celeste (https://ashleyirispup.cat)

Originally adapted from the Fedora Project AI Contribution Policy

Document Distribution & Modification: Everyone is permitted to copy and distribute verbatim copies of this document, but changing it is not allowed. Official modifications to this policy for the Poke Project may only be enacted by the project founder or through formal community consensus and voting mechanisms.

1. Definitions

In short: Clarifying what we mean by AI, automation, and the specific terms used in this policy.

  • RFC Terminology: The key words "MAY", "MUST", "MUST NOT", and "SHOULD" in this document are to be interpreted as described in RFC 2119.
  • LLM / Generative AI Coding: The automated synthesis of functional source code, scripts, or configuration files by neural network models in response to natural language prompts or broad context windows.
  • Deterministic Automation: Traditional software tools like standard static linters, CI/CD pipelines, and regex-based code formatters (e.g., Prettier, Black). These follow explicit rules rather than predictive generation and are not restricted by this policy.
  • Generative AI Assets: Any non-text media, including images, video, audio, or 3D models, synthesized by machine learning algorithms (e.g., Midjourney, DALL-E, Stable Diffusion).

2. Accountability

In short: If you submit it, you own it.

Because AI models lack legal, ethical, and technical agency, you MUST take full responsibility for your contribution. Submitting to Poke means you personally vouch for the code's quality, security, license compliance, and human authorship. All contributions must meet the project's standards for inclusion.

3. Permitted Usage and Prohibitions

In short: Starting June 3, 2026, Poke no longer accepts any new LLM-generated code. Legacy AI code stays. You may use AI strictly as a local learning or debugging tool, but not for generating submitted code.

The Generative AI Code Ban (Effective June 3, 2026)

To guarantee that contributors maintain a thorough, working understanding of the logic they submit and to protect the project's copyright integrity, the Poke Project has simplified its stance on Generative AI:

  • No New Submissions: Effective June 3, 2026, you MUST NOT submit any new source code, scripts, or configuration files generated by Large Language Models. All new logic, architectural scaffolding, and boilerplate must be individually authored by a human.
  • Legacy Code Preservation: All LLM-generated code submitted and merged prior to June 3, 2026, under previous policy versions, will remain in the repository. We will not purge historical contributions.

Permitted Local Enhancements

The Poke Project allows the use of AI tools strictly for local workflow enhancements, provided no generated code is submitted:

  • Code Explanation: Using AI locally to summarize, document, or explain complex legacy codebases to aid your understanding.
  • Reviewing & Debugging: Utilizing AI locally to analyze your human-written code for potential flaws before manually resolving them.

Strict Prohibitions

  • Code Submission: You MUST NOT submit LLM-generated code for any purpose, including tests, boilerplate, or bug fixes.
  • Generative AI Assets: You MUST NOT submit Generative AI Assets of any kind (images, video, audio, 3D models) in Pull Requests, Merge Requests, or repository assets.
  • Fully Automated Documentation: You MUST NOT submit documentation generated by GenAI. Documentation requires substantial individual authorship.
  • Proprietary Data Leakage: You MUST NOT paste proprietary, unreleased, or sensitive project data into public, consumer-grade AI models that use inputs for training.

4. Enforcement and False Positives

In short: We do not accept AI code, but we also do not trust automated AI detectors. We operate on human review.

Because all new generative code is banned, there are no longer percentage thresholds or disclosure trailers to track. However, to protect contributors from false accusations, the following rules apply:

The "False Positive" Rule

The Poke Project strictly prohibits relying on automated "AI detection" tools to enforce this policy. These tools are notoriously unreliable and frequently flag neurodivergent phrasing, non-native English writing, or highly structured coding styles as AI-generated.

  • You will never be penalized if a third-party tool inaccurately flags your work.
  • Project maintainers MUST NOT reject code, block PRs, or aggressively interrogate contributors based purely on subjective "AI vibes" or suspicions that code "looks generated."
  • Clarification: If a reviewer suspects a contributor does not understand the code they submitted, they should ask technical questions about the logic to verify human authorship and comprehension. Rejecting code simply because of its stylistic "vibe" is not permitted.

Code from External Platforms

When integrating human-written snippets from external free software repositories, wikis, or forums, the origin of the code may be unknown. Legacy code and highly optimized patterns often look identical to modern generated outputs. You are responsible for ensuring any integrated snippet was not recently generated by an LLM and that you fully comprehend its logic.

5. Community Conduct

In short: Be respectful. Focus on the code.

The Poke Project enacted these guidelines to establish clear boundaries and foster a psychologically safe environment.

  • Focus on the Code: All technical reviews, discussions, and community feedback MUST remain strictly focused on the quality, security, and technical merit of the contribution.
  • No Harassment: Community members MUST NOT harass or mock contributors regarding suspected AI usage. Accusations of AI use should be handled professionally by project maintainers through technical review, not community dogpiling.
  • Enforcement: Hostility toward contributors will be treated as a severe breach of the Code of Conduct.

6. Evaluation and Automated Approvals

In short: Automation is for testing, humans are for merging.

  • You MUST NOT use AI as the sole or final authority in making substantive or subjective decisions about a contribution's merge status. AI cannot press the "Merge" button.
  • This does not prohibit the use of Deterministic Automation systems for objective technical validation, such as CI/CD pipelines, static linters, automated testing, or spam filtering. Final accountability rests exclusively with the individual reviewer who authorizes the merge.

7. Policy Violations and Consequences

In short: Break the rules, and your PR gets closed.

Adherence to this policy is mandatory. Failure to comply with these guidelines will result in the following progressive actions:

  1. Rejected Contributions: PRs or MRs found to contain code or assets generated by AI after June 3, 2026, WILL be immediately closed.
  2. Repeated Violations: Contributors who repeatedly ignore the AI ban MAY face temporary or permanent bans from the repository.
  3. Conduct Violations: Harassment regarding AI usage is treated as a severe breach of the project's Code of Conduct and will result in disciplinary action.

8. Large Scale Initiatives & Reporting Concerns

In short: Mass-automated PR bots need special permission. Report bad behavior privately.

Large Scale Initiatives: This policy does not cover large-scale, automated deterministic initiatives that significantly change how the project operates. Such initiatives require separate, prior discussion and approval by the project council.

Reporting Concerns: Concerns about possible policy violations, including suspected AI use or AI-related Code of Conduct breaches, should be reported through appropriate private or secure project channels.