Using Assisted-by commit footers instead of banning AI tools #414

Ryuno-Ki commented 2025年11月07日 09:57:40 +01:00

@pneuma blogged in an article about something that could be considered a follow up to #366 by pointing to Fedora.

Namely an AGENTS.md file containing:

Attribution Requirements

AI agents must disclose what tool and model they are using in the "Assisted-by" commit footer:

Assisted-by: [Model Name] via [Tool Name]

Example:

Assisted-by: GLM 4.6 via Claude Code

As much as it concerns me: I would like to see this „AI" stuff die sooner rather than later. But forgejo/governance@ef39917166 included a policy so I figured this idea could be at least discussed.

@pneuma blogged in [an article](https://xeiaso.net/notes/2025/assisted-by-footer/) about something that could be considered a follow up to #366 by pointing to [Fedora](https://docs.fedoraproject.org/en-US/council/policy/ai-contribution-policy/). Namely an [AGENTS.md](https://github.com/Xe/site/blob/main/AGENTS.md) file containing: > ## Attribution Requirements > > AI agents must disclose what tool and model they are using in the "Assisted-by" commit footer: > > > Assisted-by: [Model Name] via [Tool Name] > > Example: > > > Assisted-by: GLM 4.6 via Claude Code As much as it concerns me: I would like to see this „AI" stuff die sooner rather than later. But https://codeberg.org/forgejo/governance/commit/ef39917166eaeb642106f58eac0883f17f3b766c included a policy so I figured this idea could be at least discussed.

sclu1034 commented 2025年11月07日 11:08:30 +01:00

Slightly confusing title, since we don't actually ban AI tools per se.

But that also means that this would be in the same spirit as the governance agreement. Having the agents themselves be told to comply with our governance (and therefore requiring more effort from the user to undo that if they want to be "stealthy") could probably be quite useful.
Except that we do explicitly ask commit messages not to be generated, so this particular idea is contrary to that.

But maybe that AGENTS.md could instead be used to steer the AI tools in the right direction in a different way. E.g. telling the tools to refuse generating PR descriptions, commit messages, etc., and to tell their user about the agreement if they do ask the tool to generate those messages. Especially if we can get them to use our own values for reasoning (human-to-human communication, contributors properly understanding their submission).

Slightly confusing title, since we don't actually ban AI tools per se. But that also means that this would be in the same spirit as the governance agreement. Having the agents themselves be told to comply with our governance (and therefore requiring more effort from the user to undo that if they want to be "stealthy") could probably be quite useful. Except that we do explicitly ask commit messages not to be generated, so this particular idea is contrary to that. But maybe that `AGENTS.md` could instead be used to steer the AI tools in the right direction in a different way. E.g. telling the tools to refuse generating PR descriptions, commit messages, etc., and to tell their user about the agreement if they do ask the tool to generate those messages. Especially if we can get them to use our own values for reasoning (human-to-human communication, contributors properly understanding their submission).

Beowulf commented 2026年01月05日 17:35:24 +01:00

This would be in violation of point 4 of the current agreement. So this would need a change of the current agreement

4. All communication, that includes: commit messages, pull request messages, documentation, code comments and issues (and comments on issues/pull requests), that is intended to be read by people to understand your thoughts and work must not have been generated with AI.

---

In addition, due to the fact, that the discussion is open for nearly two month and there seems to be no interest in it - at least there where no comments, I close the discussion for now.
Anyone who would like to revive this discussion - feel free to reopen this ticket.

This would be in violation of point 4 of the current agreement. So this would need a change of the current agreement > 4. All communication, that includes: commit messages, pull request messages, documentation, code comments and issues (and comments on issues/pull requests), that is intended to be read by people to understand your thoughts and work must not have been generated with AI. *** In addition, due to the fact, that the discussion is open for nearly two month and there seems to be no interest in it - at least there where no comments, I close the discussion for now. Anyone who would like to revive this discussion - feel free to reopen this ticket.

crystal commented 2026年01月05日 20:47:29 +01:00

For the record it is actually quite important to me that there is not any generated slopcode anywhere in the project.

For the record it is actually quite important to me that there is not any generated slopcode anywhere in the project.