Document AI communications policy

[bluelhf]

Problem

For some time now SkriptLang has had a de-facto rule that forbids AI authorship of maintainer—developer communications, but this is not documented in Skript's contributing guide nor in agent-specific documentation. This leads to agents, not being aware of the rule, authoring pull requests and issues.

Solution

Add a human-readable notice to .github/contributing.md that AI assistance must not be used to generate communications, as well as an agent-readable policy in AGENTS.md that instructs agents to refuse such tasks and explain the policy to the user.

Testing Completed

Installed and ran Codex on the repository asking it to make a pull request description for an innocuous change (changing the default parse time warning threshold) and confirmed that it refused.

Me: Generate a pull request description for this branch to dev/patch

Codex: I can’t generate a pull request description because this repository requires PR communications to be human-authored. AI-assisted code contributions must also be disclosed in the PR.

Supporting Information

AGENTS.md is an emerging standard that most agents will read, so it seemed to strike a balance between having a file that is very accessible to agents and not polluting the repository with useless LLM-oriented junk.

---

Completes: none
Related: all of the AI slop PRs
AI assistance: Codex to test the policy, Claude drafted the policy and gave the filepath for agent-specific documentation

[TheMug06]

lgtm otherwise :)

[APickledWalrus]

I think this reflects our stance pretty well. Thanks for this.