add agents.md

[technophile-04]

Description:

Removed the cursor rules and we now have AGENTS.md and CLAUDE.md file. Claude.md file basically references to AGENTS.md only because we don't want to maintain two separate similar copies.

1. Document requirements: Clearly specify dependencies and commands in your CLAUDE.md file. If you have an AGENTS.md file, you can source it in your CLAUDE.md using @AGENTS.md to maintain a single source of truth.
   ~ Calude Code docs

So this file will basically will be added as system promp by the harness you are using (not all but most of them do it). So for every new conversion the harness will preload this file in context.

Your CLAUDE.md file becomes part of Claude's system prompt. Every conversation starts with this context already loaded, eliminating the need to explain basic project information repeatedly.

While researching what are the best practices, there is as such no standard and pepople are actually writing plain egnlish in ths file about project. But a certain pattern I have seen in most of the famous repos is that they use this file as a guardrail for AI so that it does't go too crazy, keep the consistency for all the common things, and know how to look into the repo.

So people mainly include code style guidelines, testing instructions, repository conventions, developer environment setup, and project-specific warnings. Tried following the same flow.

I initally generated this file by using /innit command from calude code (which generated CLAUDE.md and then copied it over to AGENTS.md) and followed the pattern mentioned about.

Generally everyone suggested to keep the file max 300-600 lines and in some places I even found people suggesting to keep it under 100. Seem like 200 is sweet spot for us since we could mention all the imp things. But we can decrease and refine a bit because I feel the less liness + cripier it is better for LLM.

An nice line which I liked in Claude documentation for adding all the AI compatibility (like skills, subagents, plans.md etc) to your repo was:

Start simple, expand deliberately

[technophile-04]

I just added a couple of agents. mainly hoisting the file in .agents dir so that other agents file can reference it.

Maybe this is good enough for 0.1v iteration. I think having AGENTS.md solves alot of stuff.

Some more things we can explore / add maybe here or in other PR:

1. Having a hardhat skill, which tell how to deploy contract (Se-2 way i.e hardhat-deploy)
2. Skill for common solidity security we can have some stuff copied from eth-wingman or isntall this https://skills.sh/pluginagentmarketplace/custom-plugin-blockchain/smart-contract-security
3. A skill which we talked about merging the extensions code. So if someone says I want to build ERC-20 token we fetch the code for erc-20 extension and add integrate it with current setup of the repo.
4. Access to SRE challenges repo. Based on the given prompt it fetches the sre-challenges extension and integrates it (I think this will be solved if we solve 3 easily, since challenge are just extension)
5. We could have more frontend skills like:
   1. DaisyUI
      2. Next js skill by vercel
      3. React skill by vercel

Didn't add them here coz thought maybe they will bloat it. but we can add them in future.

[Pabl0cks]

Nice job Shiv! Added some comments, most of them suggestions that can be tackled in future PRs if we find them interesting 🙌

[technophile-04]

Update added the solidity skills took the inspiration from ethwingman and https://skills.sh/wshobson/agents/solidity-security

[rin-st]

Great job, Shiv! Thanks!

---

Let's maybe create an issue or some doc with #1221 (comment) and other potential improvements from this PR comments after merge?

[technophile-04]

Pushed some small tweak mainly including content related to foundry.

We could have used template and conditionally only show the foundry content when foundry was selected, but while discussing with Carlos we thought AI is smart enough to infer it and follow the instrction mentioned depending on flavour choose. We can also update it later to use template.

Merging this as v0.1 and we can iterate more since this doesn't have any changes to core logic of SE-2