contributions are welcome, doing too strict and legalese things are not my stile however a couple guidelines are in order.
First and foremost, be respectful, don't be toxic, write and talk in the manned you'd like to be talked to, now this sounds obvious but it may not be, altho i run things in a very informal way and recognize that internet language is what it is which is an extension of informal vulgar english try to keep your typing of uppercase "FUCK", "FUCK YOU", "PIECE OF SHIT", etc... directed at the code rather than the individuals writing it, same applies to documentation, if ya want to write your gamer words which would be at home in 2009 xbox live voice chat, league of legends chat, or just a call of duty lobby then have the full expression of your freedom of speech at a suitable platform like 4chan, the fediverse, your home, voice chat with your friends or any freeze peach platform ya can find which ain't this, this is a git forge, so unless it is code, documentation, issues or direct discussion of the software then it doesn't belong in here, that includes talks of broader politics, i could not care any less in here about what you think of bubba's glory hole, this is not the platform to verbally fellate him or call him JE's island cumrag, please go to another place for that, that ranty verbarrhea aside.
When a piece of software that is not shed is involved please be explicit, i have no way of knowing what every piece of software is from name alone, even if it seems obvious just the act of providing a link to the software's homepage in the first mention is help enough
When opening an issue please provide detailed information, not just a dry "it does not work" or a "the documentation is wrong", be specific, add information about the relevant software like names and versions, the version of shed you are using down to the commit hash (full hash is not needed), related config files for shed and the sofware involved.
Follow the 1 feature per pull request convention, ideally every pull request is related directly to 1 feature, say you want to improve logging and also add some feature that depends on the improved logging, you may feel tempted to put both of those in the same Pull Request like "improved logging and feature" but please do not, even if you got the code all ready to merge just open 2 pull requests, first "improved logging" and then "feature", in the comment for the "feature" pull request just specify it depends on "improved logging", this helps keeping the code and commit history cleaner.
On all interactions from opening issues to submitting code please follow the ai policy.
First and foremost it is important to outline, "Artificial Intelligence" is a tool, nothing more nothing less, and like every tool there will be people whom can use it to great extents and those whom completely miss the point and use it wrong be it from incompetence or malice.
For the sake of transparency and openness we classify the usage of GenAI for contributions in the following manner:
-
No AI: prospect contributions done completely by humans without the usage of large language models, agentic ai, neural networks, neural syntactical analizers, autonomous language model lexers, prompt driven code generators.
-
AI Assisted Human: prospect contributions created with the assistance of generative AI tools in which the majority of the work is still done by a human.
-
AI Co-Authored: prospect contributions created with the assistance of generative AI tools in which a similar or at least equitative part of the work was done by a human and a generative AI tool.
-
Human Assisted AI: a prospect contribution created with generative AI in which the majority of the work was done by the AI and a minority done by a human.
-
Agentic AI: a prospect contribution in which the totality of it was generated by a type of autonomous or semi-autonomous llm or neural network based system in which the human interaction is limited to prompting.
The reasons for the specific wording on these classifications is so that systems such as code completion, code template completion, language servers, lexical parsers like tree-sitter, text editors like neovim or emacs will not count as usage of AI tools as they are NOT, in the side of agentic ai the wording should suffice so that systems like debian-janitor and lintian won't be considered as agentic ai.
With the 5 classifications we got for the different levels of ai tools usage we can define the guidelines for AI usage disclosure and vouching as well as the thresholds for what enter where.
Nothing to ruminate about here, if the prospective contribution has 0% changes created through AI tools it falls in this cathegory.
Starting with this cathegory all usage of AI tools for minimal to seemingly inconsequential it may be starts at this cathegory so long as the AI tool of choice served as something more than just a fancy code template generator it goes in here, however not every usage of AI is significative enough to really bother with outspoken on front disclosure, so in being lenient i will set the threshold for disclosure of AI assistance at 1/5th or 23% (0.23, i'm being more generous than IEEE 754 on what constitues ONE FIFTH) of the changes per commit, once 23% of the changes in an individual commit come from an AI tool it is required to disclose them as "AI Assisted" and tag the ai model as a co-author in the commit message, for example if google's gemini was used the last lines of the commit message ought to look like:
AI Assisted
Co-authored-by: Gemini <gemini-code-assist@google.com>In following with the previous cathegory we consider that a commit was co-authored by an AI tool once the final percentage of changes in the commit that are authored by an AI tool surpases 1/3rd or 35% (0.35, again i'm being generous here for the sake of generosity) of the code, with an upper threshold of 65%, so once the AI provided changes in a commit hit the 36% mark but do not surpass 65% the prospective contributions shall be tagged as "AI Co-Authored" at the end of the commit message, again using google's gemini for the example:
AI Co-Authored
Co-authored-by: Gemini <gemini-code-assist@google.com>Once the amount of human contributed changes in the commit fall below 35% or in other words once the AI contributed changes surpasses the 65% threshold the commit shall have the AI tool as author, be tagged as "Human Assisted" and be co-authored by the human at the end of the commit message submitting it, in order to keep things fair and consistent for this case with the "AI Assisted Human" once the human contribution to the total of the commit's changes fall below 23%, which is 22% or less of the changes came from a human it is no longer necessary to tag the human submitter as a co-author, that said this is an example of what a commit message may look like where the human contributed around 34% of the total changes, again using gemini AI as the example AI tool:
Author: Gemini <gemini-code-assist@google.com>
Date: Sat Apr 18 15:26:07 2026 -0300
Improved lockfile management
This reduces potential race conditions.
Human Assisted
Co-authored-by: Jane Doe <jdoe@tf2mann.co>This cathegory being reserved for prospective contributions with 0% human provided changes, if the only interaction done by the human submitting it was a prompt then it befalls this cathegory, of course if the prospective contribution is being submitted by an AI agent it also befalls this cathegory, proper disclosure and vouching of the changes by a human whom can provide evidence of having tested the changes is required before the prospective contribution can be considered for merging.
Failure to comply to the AI policy will instill the following consequences:
-
Written warning: if a submitted contribution is suspected to not comply with the disclosure guidelines of the AI policy a public written warning will be issued at the pull request where the contribution was submitted asking the submitter for disclosure and clarification on AI tool usage.
-
Pull Request Closure: if the submitter fails to comply with the AI policy by refusing to disclosure evident use of AI assistance the Pull Request will be closed.
-
Removal of undisclosed code: if a piece of code was merged which is later discovered or suspected to have been AI generated or simply AI assisted but without the proper disclosure, said code will be removed and a written message issued at the pull request where the contribution was submitted.
-
Ban from Contributing: i believe in multiple opportunities to amend mistakes so this is a last resort, but prospective contributors whom continuosly fail to disclose their usage of AI tools and refuse to change their behaviour will be forbidden from contributing to the project.