feat: add AGENTS.md for AI rules

[rainerhahnekamp]

AGENTS.md seems to be an inofficial, cross-IDE standard for defining AI rules

[rainerhahnekamp]

@michaelbe812, can we work on that together. I am looking for a way on how to define coding rules independent from the IDE. As far as I know, that one https://agents.md/ comes closest.

[michaelbe812]

@rainerhahnekamp i would love to!

[michaelbe812]

@rainerhahnekamp

Not sure how I should contribute (separate PR / separate agents.md in your branch)?

Here are some first thoughts from my side:

YOU MUST FOLLOW the instructions in this document

# General guidelines
- Before starting a Task do a thorough planning and break down the task in small sub-tasks 
- Use conventional commits 
- Write atomic commits
- Before finishing a task make sure that `yarn lint:all`, `yarn test` and `yarn build:all` pass as well as the integration tests when running `run-integration-tests.sh`. To run a single test file use `yarn vitest run -t "<test name>"`
- Sheriff follows a zero-dependencies policy which means that adding a third-party package should be avoided at all costs.

# Project Overview
Sheriff is a TypeScript  tool which enforces module boundaries and dependency rules. For more information you can read the documentation in `./docs`.

## Important Packages
- `packages/core`: This is the most important package where all the logic as well as the CLI tool lives
- `packages/eslint-plugin`: ESLint Plugin using the core packages
- `test-projects`: Test projects with different setups to make sure that Sheriff works as expected. Some of the setups are covered automatically with integration tests (`integration-test.sh`)

# Implementing new features
- Actual Implementation
- Check if documentation needs to be updated/added
- JSDoc for public members
- Unit Tests (100%)
- Integration Tests

## Patterns used / Examples
tbd. Here we could list examples of common patterns to be used

# Bug Fixes

First reproduce the bug via a unit test. Once that is done, fix it in the code itself.

Not sure if agents.md also supports to list allowed tools and commands. This would also be nice to add so that the agent can work autonomously.

In general I would also describe common patterns to use in the codebase (if there are some important patterns) and always add examples wherever suitable

[rainerhahnekamp]

@michaelbe812, you can do "pair programming". So you can directly push in my branch and we discuss it here (or via a short call)

But your version above and the tips afterwards looks already much better than mine 👍

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[rainerhahnekamp]

@michaelbe812 should we merge? Should be enough for an initial version

[michaelbe812]

@rainerhahnekamp
Let's go!

It will be anyways an incremental effort to regularly update the instructions :)